United StatesDepartment ofAgriculture

Forest Service

Pacific NorthwestResearch Station

Research PaperPNW-RP-525July 2000

Smoke Exposure at Western WildfiresTimothy E. Reinhardt and Roger D. Ottmar

Timothy E. Reinhardt is a senior scientist, URS Corporation, 1500 Century Square,1501 Fourth Ave., Seattle, WA 98101; and Roger D. Ottmar is a research forester,U.S. Department of Agriculture, Forest Service, Pacific Northwest Research Station,4043 Roosevelt Way NE, Seattle, WA 98105.

Fire crew wearing smoke exposure instrument packs, Libby Complex wildfires,Montana, 1994. Photo by Roger Ottmar.

Authors

Cover

Reinhardt,Timothy E.; Ottmar, Roger D. 2000. Smoke exposure at western wild-fires. Res. Pap. PNW-RP-525. Portland, OR: U.S. Department of Agriculture,Forest Service, Pacific Northwest Research Station. 72 p.

Smoke exposure measurements among firefighters at wildfires in the Western UnitedStates between 1992 and 1995 showed that although most exposures were not significant, between 3 and 5 percent of the shift-average exposures exceeded occu-pational exposure limits for carbon monoxide and respiratory irritants. Exposure tobenzene and total suspended particulate was not significant, although the data forthe latter were limited in scope. The highest short-term exposures to smoke occurredduring initial attack of small wildfires, but the shift-average exposures were less during initial attack than those at extended (project) fire assignments because ofunexposed time during the shift. Among workers involved in direct attack of activelyburning areas and maintaining fireline boundaries, peak exposure situations could beseveral times greater than recommended occupational exposure limits for short-termexposures. The study found that exposure to acrolein, benzene, formaldehyde, andrespirable particulate matter could be predicted from measurements of carbonmonoxide. Electrochemical dosimeters for carbon monoxide were the best tool forroutinely assessing smoke exposure, so long as quality assurance provisions wereincluded in the monitoring program. Suggested procedures for reducing overexpo-sure to smoke include (1) hazard awareness training, (2) routinely monitoring smokeexposure, (3) evaluating health risks and applicable exposure criteria, (4) improvinghealth surveillance and injury recordkeeping, (5) limiting use of respiratory protectionwhen other mitigation is not feasible, and (6) involving workers, managers, and regu-lators to develop a smoke exposure management strategy.

Keywords: Smoke exposure, firefighters, occupational health, pollutants, safety,industrial hygiene, smoke hazards.

Abstract

This project measured smoke exposure among wildland firefighters in the WesternUnited States between 1992 and 1995. The objectives of the study were to assessfirefighter exposure to air pollutants in smoke at wildland fires in the Western UnitedStates, determine the average and variability of smoke exposure among wildland firefighters, and observe factors that controlled the exposures. We evaluated the performance of recently available tools, such as dosimeters, to measure smoke expo-sure. Finally, we determined whether pollutants in smoke were sufficiently correlatedwith one another for measurements of one pollutant to be used as a surrogate to estimate exposure to the others.

Breathing-zone measurements of acrolein, benzene, carbon dioxide (CO2), carbonmonoxide (CO), formaldehyde, and particulate matter (total and respirable) wereobtained concurrently during active firefighting by using personal sampling pumpsand sampling media worn by the firefighters. Electrochemical dosimeters also wereused to measure CO, thereby providing the advantage of continuous exposurerecords. Over 1,750 separate measurements of pollutant exposure were collectedand analyzed by the project laboratory of the U.S. Department of Agriculture, ForestService, Pacific Northwest Research Station. The results were validated by an extensive quality assurance program.

Smoke exposure measurements were made at 13 wildfires in California that weresuccessfully controlled by initial attack forces (referred to as “initial attack wildfires”),and during eight multiday wildfires (“project wildfires”) in California, Idaho, Montana,and Washington, for a total of 30 days of firefighting. For the most part, hand crewswere monitored at project wildfires and engine crews at initial attack wildfires. Thedata from the initial attack wildfires and project wildfires were analyzed separatelybecause firefighters at project wildfires usually spend more time at a fire than do initial attack forces, thus the average smoke exposures per shift (“shift-averages”) are very different.

Based on our experience and the observations of experienced firefighters we spokewith, the conditions we monitored were typical of most days of wildland firefighting in the Western United States. We say “most” because we have observed other fires(and heard about more from experienced firefighters) where conditions seemedsmokier than those we successfully sampled, by both visual assessment and person-al adverse health effects. At the fires we were able to sample, we found that mostsmoke exposures would not be considered hazardous, but a small percentage routinely exceeded recommended exposure limits at project wildfires. Only CO wasshown to occasionally exceed full-shift permissible exposure limits (PELs), as estab-lished by the Occupational Safety and Health Administration (OSHA), because theCO exposure limit is adjusted downward by a reduction factor proportional to the shiftlength, which averaged 14 hours at project wildfires. About 3 percent of the firefight-ers’ exposures exceeded the adjusted OSHA PEL for CO at project wildfires. None ofthe exposures exceeded adjusted PELs at initial attack wildfires. The total respiratoryirritant exposure (a combination of exposure to acrolein, formaldehyde, and respirableparticulate matter) approached the PEL while firefighters were working at projectwildfires, but unexposed travel time within the work shifts served to bring the shift-average exposures below OSHA limits. Based on our results, firefighting agenciesmay want to consider adopting the Threshold Limit Values as occupational exposurecriteria (guidelines recommended by the American Conference of Governmental

Summary

Industrial Hygienists [ACGIH]), which incorporate more current knowledge of healtheffects than the Federal OSHA PELs, some of which are based on limited data fromthe late 1960s. About 3 percent of the shift-average respiratory irritant exposures andabout 5 percent of the CO exposures at project wildfires exceeded the recommendedACGIH Threshold Limit Values.

The shift-average exposures to smoke at project wildfires were greater than at initialattack wildfires, but smoke exposure was generally greater during suppression at ini-tial attack wildfires than at project wildfires. The lower smoke exposures among initialattack firefighters at wildfires were due to long periods of unexposed time bracketingtheir brief exposures during active suppression, thus bringing their shift-averageexposures into compliance with both PELs and Threshold Limit Values. Exposure tobenzene and CO2 in smoke was not shown to be a health concern, but the resultsshowed benzene exposure levels were consistently higher among firefighters workingwith gasoline, such as sawyers and operators of water pumps and drip torches.Exposure to total suspended particulate was within the PELs, but exposure to totalsuspended particulate was not well characterized because relatively few sampleswere collected and none were from the higher smoke conditions that occurred during1992-94. Because of this, future research direction may include further assessmentof total suspended particulate exposure as well as exposure to crystalline silica, anuncharacterized hazard that could be cost-effectively assessed through chemicalanalyses of archived particulate samples.

Samples obtained during peak exposure situations in heavy smoke indicate thatthese brief but intensely smoky situations are the driving factor behind most shift-average overexposures to smoke, although inversion conditions can cause overexpo-sures at large fires in confined topography. We found that peak exposure samplesexceeded the recommended ACGIH short-term exposure limits for the mixture of respiratory irritants (acrolein, formaldehyde, and respirable particulate matter) inabout 50 percent of our peak exposure measurements.

Some differences in smoke exposure were apparent among different job tasks:smoke exposure was lower during mop-up and greater during line holding (whendownwind or uphill of the fire) and during direct attack of spot fires and flanks of wild-fires. Smoke exposures during initial attack were generally higher than at projectwildfires, possibly because the firefighters are, in effect, in emergency situations dur-ing the former and willing to endure short-term smoke exposure to meet initial attackgoals. Increasing ambient windspeeds were positively correlated with higher smokeexposure during fire attack tasks. The firefighters with the highest shift-averagesmoke exposures were invariably working in peak exposure situations during part oftheir work shifts. This very important finding indicated that management actionsdirected at these peak smoke exposure situations will control shift-average smokeexposures as well.

Several results of the study hold promise for minimizing future smoke exposure management costs:

• The known pollutants in smoke were highly correlated with each other. The excep-tions were total suspended particulate (because it includes entrained soil dust aswell as particles from smoke) and benzene (because firefighters may be working

with other sources of benzene, such as gasoline). At this time, respiratory irritants(acrolein, formaldehyde, and respirable particulate matter) and CO are the onlywell-documented smoke hazards to firefighters. Monitoring for CO therefore canmeet most monitoring needs through the use of correlations to estimate exposureto the respiratory irritants.

• Visual estimates hold promise for a low-cost method of determining opportunitiesfor smoke management action. The visual smoke exposure classifications madeby an observer with moderate fire experience correlated fairly well with actualsmoke exposures.

• The electronic CO dosimeter can give accurate, instantaneous warnings to person-nel when CO levels (and correlated respiratory irritants) exceed predefined limits.With data-logging capabilities, this type of instrument can form the basis for awidespread smoke exposure monitoring program with the added advantage ofdirect feedback to fireline personnel.

If the goal of firefighting agencies is to reduce smoke intake by firefighters becausethe highest smoke exposures exceed recommended and permissible exposure limits,the management program could include the following elements:

• Hazard awareness training to help managers and firefighters understand why andwhen smoke exposure is likely to be a concern, what chemicals and physicalagents in smoke are involved, how to recognize symptoms of overexposure, andhow to manage work to reduce smoke exposures.

• Smoke exposure monitoring, mainly using electronic CO dosimeters (supplement-ed by occasional comprehensive monitoring), to provide instant feedback to firepersonnel, routinely check compliance with occupational exposure limits and evalu-ate the effectiveness of smoke exposure management strategies by tracking trendsin smoke exposure data over many years.

• Evaluation of exposure limits and health risks among wildland firefighters, toassess short- and long-term health consequences and to evaluate whether expo-sure limits developed for standard industrial workplaces and career patterns areadequate for firefighters.

• Improved recordkeeping of health data to quantify whether smoke-related injuriesand illnesses occur, track trends in firefighter health, and enable the assessment oflong-term health risks among firefighters.

• Limited respirator use (in accordance with a respiratory protection program and COmonitoring) to reduce respiratory irritant exposure and allow highly trained person-nel to work effectively for brief periods in heavy smoke.

• Involvement and consensus building among managers, workers, and regulators todevelop a workable smoke exposure management program and maintain continu-ous improvement.

1 Introduction

1 Background

1 Health Hazards in Smoke

3 Smoke Exposure Evaluation Criteria

4 Multiple Contaminants

4 Exposure Duration Considerations

5 Objectives

6 Methods

6 Field Methods

9 Laboratory Methods

10 Quality Assurance Program

11 Method Detection Limits

11 Data Analysis

12 Pollutant Correlations

12 Time-Weighted Average Exposures

13 Peak Exposures

14 Job Task

15 Environmental Variables

15 Dosimeters

15 Observer Estimates

17 Results

17 Data Quality

19 Pollutant Correlations

23 Exposure Assessment for Project Wildfires

28 Exposure Assessment for Initial Attack Wildfires

Contents

34 Peak Smoke Exposure Assessment

35 Factors Influencing Smoke Exposure

43 Dosimeter Performance

44 Observer Estimates

48 Discussion

50 Pollutant Correlations

52 Exposure Assessment

52 Exposure Assessment for Project Wildfires

56 Exposure Assessment for Initial Attack Wildfires

58 Factors in Smoke Exposure

63 Dosimeter Performance

65 Observer Smoke Estimates

65 Management Implications

66 Training in Hazard Awareness

66 Monitoring of Routine Smoke Exposure

66 Evaluating Exposure Limits

67 Improving Recordkeeping

67 Assessing Health Risks

67 Deploying Respirators

67 Involving Workers and Regulators

68 Abbreviations

69 Acknowledgments

69 References

1

This paper summarizes measurements of smoke exposure among wildland firefighters in the Western United States. Smoke exposure measurements were taken from over129 firefighters (both hand crews and engine crews) at 13 initial attack and eight proj-ect fires in California, Idaho, Montana, and Washington between 1992 and 1995.

Wildland firefighting presents many hazards to firefighters, including burns, heatstress, tripping and falling hazards, accidents with hand and power tools, being struckby falling rocks and trees, encountering poisonous insects, reptiles, and plants, andinhalation exposure to smoke. Many experienced firefighting personnel consider thelast to be only an inconvenience that occasionally causes acute eye and respiratoryirritation, nausea, and headache. Others express concern about longer term healthimpacts, especially when large-scale fires occur in terrain and atmospheric conditionsthat force firefighters to work for many days in smoky conditions.

Support and coordination of smoke exposure research and information has been pro-vided by the National Wildfire Coordinating Group (NWCG),1 including an informativequarterly newsletter, “Health Hazards of Smoke,” which received wide distribution fromthe U.S. Department of Agriculture, Forest Service, Missoula Technology DevelopmentCenter (MTDC). Several studies of smoke exposure have been undertaken by fire -fighting agencies since the early 1970s. Since the late 1980s, the National ParkService (NPS) has funded several studies by the National Institute for OccupationalSafety and Health (NIOSH). As well, the Pacific Northwest Region of the U.S. Depart-ment of Agriculture (USDA), Forest Service, and the Bureau of Land Management(BLM), the California Department of Forestry and Fire Protection (CDF), and NWCGfunded the USDA Forest Service, Pacific Northwest Research Station, to conduct apreliminary assessment of smoke exposure at wildfires in the Northwestern UnitedStates, which was completed in 1995 (Reinhardt and others 1995b). The USDAForest Service and CDF also supported another assessment of smoke exposureamong initial attack crews in Redding, California, also completed in 1995 (Reinhardtand others 1995a). An overview and summary of some of the relevant literature maybe found in Reinhardt and Ottmar (1997a).

This research paper combines additional smoke exposure data obtained in the 1995wildfire season with the data from the previous two studies for a more completeassessment of smoke exposure at wildfires. It examines smoke exposure in the earlystages of fire suppression (initial attack) and during extended attack at project wild-fires. Although the results are mainly representative for much of the Western UnitedStates, similar results are possible in other regions. Situational factors (such as wherethe firefighter is in relation to the fire, the ambient wind speed, the fire behavior, theterrain and fuel burning, and the urgency of the work task) ultimately determinewhether smoke exposures will be greater or less than what we report here.

Smoke from wildland fires is composed of hundreds of chemicals in gaseous, liquid,and solid forms. The chief inhalation hazards seem to be carbon monoxide (CO),aldehydes, respirable particulate matter with a median diameter of 3.5 micrometers(PM3.5), and total suspended particulate (TSP). Many low- to middle-molecular-weight aldehydes are present in smoke, but formaldehyde and acrolein have been themost studied. Benzene (C6H6) is present in wildland fire smoke, but earlier work

Introduction

Background

Health Hazards inSmoke

1 All abbreviations are given in a section, “Abbreviations,” onpage 67.

shows that it is not likely to be hazardous unless poor work practices cause exposureto gasoline vapors during engine operations or drip-torch work activities (Reinhardtand others 1995b). Finally, although carbon dioxide (CO2) is present at relatively highlevels in smoke, it is diluted to nontoxic levels by the time it reaches firefighters.

Adverse health effects of smoke exposure begin with acute, instantaneous eye andrespiratory irritation and shortness of breath but can develop into headaches, dizzi-ness, and nausea lasting up to several hours. The aldehydes and PM3.5 cause rapidminor to severe eye and upper respiratory tract irritation. Total suspended particulatesalso irritate the eyes, upper respiratory tract, and mucous membranes, but the largerparticles in TSP do not penetrate as deeply into the lungs as the finer PM3.5 particlesdo. Longer term health effects, lasting days to perhaps months, have recently beenidentified among wildland firefighters, including very small losses of pulmonary func-tion—unnoticeable in most. These small decrements—on the order of a few percent-age points—include a slightly diminished capacity to breathe, constriction of the respiratory tract, and hypersensitivity of the small airways (Letts and others 1991, Reh and others 1994, Rothman and others 1991).

A discussion of particulate inhalation hazards faced by firefighters is incomplete without mentioning crystalline silica, which can be a hazard in the absence of smoke.Because crystalline silica is present in soil dust in many areas of the United States, itis likely to be a component of TSP, but it is unlikely to be significant in PM3.5 becausethe combustion products dominating the PM3.5 portion of TSP are generally low in sil-ica. Chronic exposure to crystalline silica is well-known to cause silicosis (thickeningof the lung tissues), which limits breathing ability. Only one study has attempted tomeasure crystalline silica in the dust exposures among firefighters, and that studyfound measurable levels of silica in 25 percent of the samples, ranging up to 9 percentby weight, although the highest sample was just below the Occupational Safety andHealth Administration (OSHA) permissible exposure limit (Harrison and others 1992).

Carbon monoxide causes acute effects ranging from diminished work capacity to nausea, headache, and loss of mental acuity. It has a well-established mechanism of action, causing displacement of oxygen from hemoglobin in the blood and affectingtissues that do not stand loss of oxygen very well, such as the brain, heart, and inpregnant firefighters, the fetus. Fortunately, most of these effects are reversible and

2

Table 1—Selected U.S. occupational exposure limits in 1997–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Carbon ParticulateExposure limit Acrolein Benzene monoxide Formaldehyde matter–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

- - - - - - - - - - - - - - - - - Parts per million - - - - - - - - - - - - - - - Mg/m3

OSHA permissibleexposure limit 0.1 TWA 1.0 TWA 50 TWA 0.75 TWA 5 TWA

5.0 STEL-C 2.0 STEL (respirable)15 TWA (total)

NIOSH recommended .1 TWA .1 TWA 35 TWA .016 TWAexposure limit .3 STEL 1.0 STEL-C 200 STEL-C .1 STEL-C

ACGIH threshold .1 TWA .5 TWA 25 TWA .3 TWA-C 3 TWAlimit value .3 STEL 2.5 STELa (respirable)

CAL-OSHA .1 TWA 1.0 TWA 35 TWA .75 TWA 5 TWA.3 STEL 5.0 STEL-C 200 STEL-C 2.0 STEL (respirable)

–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––aProposed in ACGIH 1996.

CO is rapidly removed from the body, with a half-life on the order of 4 hours. Somestudies have linked CO exposure to longer term heart disease, but the evidence is not clear cut.

This research paper summarizes measurements of exposure to acrolein, benzene,CO, CO2, formaldehyde, PM3.5, and TSP. The other chemicals in smoke seemunlikely to pose a significant health hazard, based on current knowledge. We cautionthat this conclusion may change as knowledge develops about toxicology and smokeexposure (Dost 1991).

To evaluate occupational exposures to chemicals and other airborne hazards, the linebetween safe and unsafe exposures requires careful evaluation because individualsdiffer in their susceptibility to adverse health effects. The OSHA sets legally enforce-able permissible exposure limits (PELs) for the United States, and many states havean equivalent occupational safety agency, such as CAL-OSHA in California. Theseagencies have established PELs that are at least as stringent as OSHA’s and may bemore so. The PELs have been established for only a small percentage of hazardouschemicals. Where PELs have not been established or are inadequate, OSHA and thestate agencies have the authority to require employers to provide a workplace freefrom recognized hazards likely to cause serious harm.

The procedures for establishing PELs are time consuming and costly, so voluntaryguidelines help to fill this gap. One source of current information is NIOSH, whichadvises OSHA on health hazards in the workplace and establishes recommendedexposure limits (RELs) that are based on detailed scientific information. Similarly, theAmerican Conference of Governmental Industrial Hygienists (ACGIH) annually pub-lishes a well-regarded compendium of Threshold Limit Values (TLVs) for worker safety.

Exposure limits for airborne pollutants are defined via three basic time categories.These categories of exposure limits are:

• Time-weighted average (TWA), a concentration for a normal 8-hour day in a 40-hourworkweek, to which nearly all workers may be exposed for a working lifetime withoutadverse effect.

• Short-term exposure limit (STEL), a maximal concentration to which workers can becontinuously exposed for up to 15 minutes without adverse effect.

• Ceiling limit (C), a concentration that is unsafe to exceed, for over 1 minute.

A "skin" notation for a chemical indicates that dermal absorption is an important expo-sure route to consider when developing management strategies. Table 1 summarizesthe current regulatory exposure limits for hazards faced by firefighters and the mostcurrent (or proposed) guidelines recommended by some key occupational healthorganizations.

We consider the TLVs to be the best starting point for assessing exposures becausethey incorporate the latest scientific evidence, whereas the PELs do not. We refer tothe ACGIH TLVs as our key evaluation guidelines in this report, although other stan-dards may be appropriate.

3

Smoke ExposureEvaluation Criteria

In a situation such as firefighting, where workers face multiple air pollutants, it is pru-dent to consider the combined effects of the pollutants. Acrolein, formaldehyde, andrespirable particulate all cause irritant effects in the same organs: the respiratory tractand mucous membranes. Beyond the physical irritation caused by fine particles,chemical analyses of woodsmoke particulate have shown it to be composed of manyorganic compounds, some of which are chemical irritants. Without detailed knowledgeof the chemical composition of the particulate, it is reasonable to assume that expo-sure to PM3.5 and the aldehydes produces an additive irritant effect in the respiratorytract and mucous membranes of the eyes, nose, and throat. Following the approachrecommended by ACGIH and OSHA, a combined "equivalent irritant exposure" indexcan be calculated from equation (1):

where Em = the equivalent exposure irritant index (a ratio);

conc. = the measured concentration of the irritant;

[C3H4O] = acrolein (ppm);

limit = the adopted exposure limit of the irritant—the PEL or TLV;

[HCHO] = formaldehyde (ppm); and

[PM3.5] = respirable particulate (mg/m3).

The total irritant exposure was an important exposure evaluation criterion in thisreport. The equivalent exposure (Em) is required to be maintained below 1.0 for aworkplace to be considered in compliance with the criterion. We evaluated the Emin two ways, by using first the recommended TLVs and then the OSHA PELs as thedivisor for each pollutant's exposure. If considering different exposure limits, simplysubstitute the appropriate exposure limits in the denominators of the calculation toevaluate the combined exposure. Our estimate of Em probably underestimates theirritant nature of smoke because we have no detailed exposure data for all the irritants(such as other aldehydes, formic acid, and possibly sulfur dioxide). If exposure datafor other irritants become available, it may be appropriate to include them in anexpanded equation. We considered other common effects of the pollutants, such ascarcinogenicity for benzene and formaldehyde, but at this time it is not suggested thatthese two chemicals be added as an Em for cancer effects because they affect differ-ent organs.

Exposure limits are developed to prevent adverse health effects that occur above acertain dose of pollutant. Dose is the amount of pollutant delivered to a target organand depends on the level and duration of exposure as well as the rates of pollutantuptake and elimination by the body. Wildland firefighting often requires hard physicallabor over extended work shifts. The exposure duration and rate of pollutant uptakein these conditions differ from the assumptions for traditional industrial workplacesused to develop occupational exposure criteria. To account for such differences,

4

MultipleContaminants

Exposure DurationConsiderations

conc.(C3H40)___________limit (C3H40)

conc.(HCHO)___________limit (HCHO)

conc.(PM3.5)___________limit (PM3.5)[ [ [ ]]]+ +Em = , (1)

adjustments are required to maintain the peak dose below the level that workerswould experience in a standard workplace in compliance with the exposure limit.Very complicated models exist to predict doses in given exposure regimes for somepollutants, such as the recently modified Coburn-Forster-Kane (CFK) equation for CO(Smith and others 1996). The simplest approach is to compare the pollutant level dur-ing the most exposed 8 hours in the day with the existing standard, while consideringthe exposure during the rest of the work shift. An alternative and objective method is to multiply the standard exposure limit by a reduction factor to achieve equivalentprotection in the nonstandard work environment.

To evaluate compliance of nontraditional work shifts with 8-hour PELs, OSHA usesone of two simple formulas to calculate an exposure limit reduction factor. Knowledgeof the toxic effects of the pollutant is needed to assign the pollutant to one of six "workschedule categories,” which then determines the correct formula to use (U.S. Depart-ment of Labor 1979). Adjustments are not always suggested for all pollutants. Expo-sure limits for acrolein and PM3.5 are based on the acute irritant effects and do notrequire adjustment downward for longer work shifts . The ACGIH formaldehyde limit of 0.3 parts per million (ppm) is a ceiling limit, intended to protect most of the workforce from irritant effects (American Conference of Governmental Industrial Hygienists1996); therefore, it does not need to be adjusted downward. The benzene TLV isbased on systemic effects, so the exposure limit can be adjusted to account for anincreased work shift duration. The CO exposure limit is based on acute toxic systemiceffects; thus adjustment is often suggested for work shifts of longer duration.

Equation (2) shows the recommended model from OSHA:

adjusted CO exposure limit = PEL 38

——— duration

,(2)

whereadjusted CO exposure limit = the revised exposure limit to account for the

extended work shift,

PEL = the permissible exposure limit (or other exposure limit, such as the threshold limit value), and

duration = the duration of the extended work shift (hours).

Although the OSHA model is straightforward and will be used in this report, the CFKequation is a better model for adjusting the CO exposure limit if detailed exposure,pulmonary function, and site altitude data are available.

The objective of this research was to provide useful information about the occurrenceand significance of smoke exposure among firefighters at wildland fires. Through bet-ter understanding of the extent, magnitude, and reasons for overexposure to smoke,cost-effective and workable measures can be developed to manage the problem.To meet this overall objective, our goals were to:

5

Objectives

• Measure firefighters’ exposure to the important contaminants in smoke at enoughfires to estimate their shift-average smoke exposure at wildfires and to determinethe variability of these exposures.

• Observe the work activities, fire behavior, and environmental factors that contributeto high smoke exposures to allow development of targeted exposure managementactions.

• Evaluate some techniques, such as CO dosimetry and visual smoke estimating,whereby fire managers could routinely characterize smoke exposure as it occurs,thereby enabling objective assessments of smoke hazards and application of appropriate protective measures.

• Find out whether the ratios among the important hazards in smoke are consistentenough that measurements of a single indicator pollutant could serve as a surrogateto estimate exposure to the rest. If these interpollutant correlations were strongenough, the future costs of routine smoke exposure monitoring could be greatlyreduced.

Data collection took place at two types of wildfires: initial attack wildfires that weresuccessfully suppressed within hours of starting, and project wildfires that took longerto control. Two key differences were considered likely between the two types of wild-fires. First, the initial attack crews work at a faster pace for a shorter time becausethey are in an emergency-response situation; project wildfire crews often take a moremeasured approach but over a longer timeframe. Second, most of the work shift atproject wildfires is spent at the fire, but most of the work shift at initial attack wildfiresis spent waiting for a fire to occur. Because of these differences, we expected verydifferent exposure patterns between the two milieus. Thus, the exposure data for initial attack wildfires and project wildfires were analyzed separately. The field andlaboratory methodology was essentially the same at both types of wildfires.

Sampling at project wildfires included several types of crews. Most were type I (hot-shot) or type II hand crews and included Native American crews, contractor crews,and Federal and state agency crews. Along with the hand crews, two engine crewswere monitored at the project wildfires. At the initial attack wildfires, both ForestService hotshot crews and CDF engine crews were monitored.The CDF crews werebased at a fire station in Redding, California, and the Forest Service hotshot crewswere from the Angeles National Forest in southern California.

Project wildfires were selected in the Western United States based on logistical feasi-bility, potential for smoke exposure, and convenience; we opted for fires that werenearby or appeared more difficult to control over fires that seemed unlikely to last verylong. During weeks when monitoring was planned at initial attack wildfires, crewswere preselected for the study based on their proximity to high fire hazard areas andtheir willingness to cooperate. Once selected, the initial attack crews were normallydispatched and sampling was done at every daytime fire that occurred.

6

Methods

Field Methods

Our sampling strategy was intended to:

• Select fireline personnel who were representative of crews expected to encountersmoke.

• Maximize the number of firefighters monitored each day to improve the likelihoodof including highly exposed personnel.

• Provide data from which smoke exposure can be assessed versus fire characteris-tics and work activities.

• Assess interpollutant correlations.

• Obtain peak exposure assessment data during brief but intense smoke exposuresituations.

• Provide field quality assurance data.

Between two and six firefighters were selected for exposure assessment at each fire,usually by asking for volunteers from within a single fire crew. Although smokers andnonsmokers were included in the sample pool, smokers were asked to refrain fromsmoking (and generally did) while the samples were being collected. The study teamcoordinated tracking the firefighters, set up and calibrated sampling equipment, andrecorded observations about smoke intensity and job task for each firefighter.

Firefighters selected for monitoring worea 4-kilogram sampling apparatus,shown in figure 1. The apparatus con-sisted of three battery-powered person-al sampling pumps held in a web-gearpack or rucksack. The three samplingpumps for a given firefighter were oper-ated concurrently during each samplingperiod, with each pump dedicated toseparate sample collection media:

• An inert gas sampling bag for collec-tion (at a fixed rate between 20 and200 milliliters/minute) and later analy-sis of CO and CO2 by IntersocietyCommittee Method (ICM) 128 (Lodge1989).

• Sorbent tubes for collection (on char-coal at 0.15 liter/minute) and lateranalysis of benzene according toNIOSH method 1501 (NationalInstitute for Occupational Safety and

7

Figure 1—Smoke exposure sampling apparatus.

Health 1989c), and formaldehyde and acrolein (on dinitrophenylhydrazine-coated C-18/silica gel at 0.2 liter/minute) according to EPA method TO-11 (U.S. Environ-mental Protection Agency 1986).

• A Teflon® 37-millimeter, filter cassette and nylon cyclone assembly for sample collec-tion and later analysis of respirable particulate matter at 1.7 liters/minute accordingto NIOSH method 0600 (National Institute for Occupational Safety and Health1989b).2 In 1995, sampling also was done for total suspended particulate using only the filter-cassette assembly and according to NIOSH method 0500 (NationalInstitute for Occupational Safety and Health 1989a).

In 1995, the gas sampling pump and bag system (for CO and CO2 analysis) was discontinued in favor of an electronic data-logging dosimeter measuring only CO,according to OSHA method ID-209 (U.S. Department of Labor 1993). The other major change in 1995 was the initiation of monitoring for total suspended particulate.

The air samples were collected consecutively from the breathing zone of each fire-fighter during discrete time periods of their work shift. Each sample period lasted forthe duration of a particular job task, or until the sample media—such as gas samplebags—approached capacity. Peak exposure (STEL) samples lasted for the duration ofthe peak exposure situation; typically, they were obtained over 15 minutes to ensurethat a sufficient sample was acquired for analysis. Each day at project wildfires,sample durations were set up to be about 2 hours, supplemented by STEL samples inobvious peak exposure situations. Conversely, at initial attack wildfires, sampling wasplanned to begin with STEL samples and switch to 1- or 2-hour samples as the firewas controlled and conditions became less dynamic. After the conclusion of a sampleperiod, a new sample was begun as quickly as possible if smoke exposure continued,but we often did not sample while firefighters were in smoke-free air, thereby allowingus to minimize unnecessary sampling and prevent dilution of CO sample concentra-tions below detection limits for method 128. Sampling usually did not begin untilsmoke reached the firefighter. If a firefighter took a work break in clean air or movedout of smoke, sampling often stopped (or paused) and resumed when smoke expo-sure continued. These clean-air situations often comprised a large portion of the day.Air pollutant exposures during these portions of the day were estimated as equivalentto background levels for the TWA calculations.

Some bias is possible in the results; for example, only volunteers were monitored, yet uncooperative workers who see no harm in smoke exposure may endure highersmoke exposure than volunteers. Another potential source of bias is that crew super-visors may have managed the activities and smoke exposure of their personnel differ-ently when the study team was present. Finally, the added weight of the samplinggear could have diminished the firefighter’s work, but as most firefighters were orga-nized into tightly knit squads, the monitored firefighters were unlikely to work at a dif-ferent pace than the rest of their crew.

The electronic CO dosimeters, along with passive sorbent tube dosimeters for CO,were evaluated for accuracy and practicality in the field during 1992-94. The passivesorbent tubes were not found to be practical in the heat and harsh conditions of

8

2 The use of trade or firm names in this publication is for readerinformation and does not imply endorsement by the U.S.Department of Agriculture of any product or service.

wildland firefighting. The dosimeters performed adequately enough that they wereused as the primary CO data collection method in 1995 after a complete field qualityassurance (QA) program was developed for them.

Analytical methods for each pollutant are outlined in this section. For a detailed discussion of each method, refer to the standard operating procedures (SOPs),appendices C through J in the final report (Reinhardt and Ottmar 1997a).

Acrolein and formaldehyde—Minor modifications were made to U.S. EnvironmentalProtection Agency (EPA) method TO-11 to analyze concentrations of formaldehydeand acrolein. This method combines aldehydes and ketones in smoke with 2,4- dinitrophenylhydrazine (DNPH) in the presence of acid to form stable DNPH deriva-tives. Derivatives were formed during sampling by drawing the air sample through asorbent tube impregnated with acidified DNPH solution. The sorbent tube (Sep-Pak®)was extracted with acetonitrile in the field to yield a solution of aldehyde-DNPH deriva-tives. Limiting pre-extraction storage improved acrolein recoveries. The solutionswere analyzed in the laboratory by high-performance liquid chromatography (HPLC)with ultraviolet detection.

This method was improved upon for acrolein. Method TO-11 specifies recoveries aslow as 40 percent for acrolein. The acrolein-DNPH derivative has a reactive doublebond, which causes losses of the acrolein-DNPH derivative immediately upon sam-pling and during storage. We have found that acrolein-DNPH degrades to severalunknown "X-acrolein" DNPH breakdown products, similar to a report by EPA research-ers (Tejada 1986). Difficulties with chromatographic resolution masked this problemearly in the project. Once good resolution was achieved, the problem and the impos-sibility of quantitative recovery data became apparent; acrolein data for the 1992 fieldseason therefore were unusable. Chromatography improvements enabled the resolu-tion and measurement (as acrolein) of the acrolein-DNPH degradation products.Quantitative recoveries of known amounts of acrolein spiked on media were regularlyachieved with the revised method.

Benzene—A slight modification to NIOSH method 1501 was used for benzene, inwhich the sampling pump pulled smoke through a small glass tube containing twosections of granular activated charcoal. Hydrocarbons such as benzene wereadsorbed on the surface of the front charcoal section. The charcoal tubes wereopened and each section desorbed with carbon disulfide (CS2). The CS2 extract wasanalyzed for benzene by gas chromatography with flame ionization detection. Ourmodification to the method used large-capacity sorbent tubes to prevent benzenebreakthrough into the back section of the tube.

Carbon monoxide and carbon dioxide—Both CO and CO2 were sampled prior to1995 by filling inert gas sampling bags via a pump. Gas bag samples were analyzedby nondispersive infrared spectroscopy using ICM 128.

In 1995, data-logging electronic dosimeters were used as the primary CO measure-ment method, under OSHA method ID-209. Laboratory aspects of this method werelimited to routine calibrations of the instruments and manipulation of data files. Zerodrift in the data-logger results was occasionally observed down to -2 ppm. Thesewere corrected manually by adding a corresponding constant (up to +2 ppm) to allaffected results.

9

Laboratory Methods

Respirable and total suspended particulate—Respirable particulate was selectivelysampled from smoke by drawing the air sample through a cyclone device at 1.7 litersper minute and collecting the PM3.5 on a filter, according to NIOSH method 0600.The total suspended particulate method (NIOSH 0500) used in 1995 was essentiallythe same but omitted the cyclone in the sampling train. The main modification to boththese NIOSH methods was that Teflon filters with a 2.0 micrometer (µm) pore sizewere used to slightly improve capture efficiency for the small particles in smoke andfacilitate future chemical analyses of the filters.

A comprehensive QA program was developed for the project by an independent QAofficer not otherwise involved in the project. The data collection effort took placeunder a QA project plan (Radian Corporation 1993) and its subsequent revision(Radian Corporation 1996). The QA program required calculation and evaluation ofmany quality control (QC) parameters obtained from laboratory and field QC samplesand assessment procedures. All QC results were recorded on control charts, andtrends were assessed throughout the project. This QA program enabled identificationand correction of many problems before they affected the quality of the data. Chain-of-custody records were maintained from sample collection through laboratory analysis.Comprehensive field and laboratory audits of the project were conducted semiannual-ly by the QA officer. Results of these audits and a complete QA review of the data arepresented in detail in appendix H of the final report (Reinhardt and Ottmar 1997a).

Several different types of QC samples were obtained at each fire to assess the vari-ability and accuracy of field data and meet the QA objectives. Systematic problemsaffecting sample accuracy were tracked through trip blanks, field blanks (unsampledmedia), field method spikes, and field matrix spikes (media spiked with knownamounts of the target pollutants, that were either analyzed directly or used to samplesmoke next to unspiked media to determine recovery of the spike). Precision of fielddata was assessed with field replicates (multiple adjacent samples of smoke in ambient air).

10

Quality AssuranceProgram

Table 2—Method detection limits for pollutants measured at wildfires––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Method detection limit–––––––––––––––––––––––––

Pollutant Method STEL samplea TWA sampleb

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––- - - - - - Parts per million - - - -

Benzene NIOSH 1501 0.032 0.004Acrolein EPA TO-11 .024 .003Formaldehyde EPA TO-11 .048 .006Carbon monoxide ICM 128 .6 .6Carbon monoxide OSHA ID209 1.7 1.7Carbon dioxide ICM 128 7.6 7.6

--Milligrams per cubic meter--Respirable particulate matter NIOSH 0600 0.935 0.117Total suspended particulate NIOSH 0500 .549 .069––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––aNominal sample duration of 15 minutes.bNominal sample duration of 2 hours.

As in the field, QC samples and parameters were used to assess and maintain controlin the project laboratory’s analytical and data manipulation systems. Blanks and inde-pendent calibration checks were used to establish accuracy, and duplicate analyseswere used to track analytical precision. The QA plan also included routine evaluationof instrument stability and performance by assessing key calibration parameters.Finally, independent blind (unknown concentration) performance evaluation sampleswere prepared by outside laboratories and submitted to the USDA Forest Service,Pacific Northwest Research Station, project laboratory—annually for difficult methodssuch as TO-11.

Method detection limits (MDLs) were periodically evaluated for each analytical methodto define the lowest concentration measurable with 99-percent confidence that it wasgreater than zero. For the first 2 years of the project, permeation tubes and a dilutionsystem were used to generate known atmospheric concentrations of each chemical ina sampling manifold. Replicate samples were then obtained from the manifold byusing the field sampling protocol. These sample results were used to determine theMDLs under EPA procedures (U.S. Environmental Protection Agency 1984). Similartests were done by using National Institute of Standards and Technology (NIST) gasesfor CO and CO2 and eight simultaneous low-level field replicate samples from one firefor respirable particulate. By the 1995 season, all MDLs were determined from low-concentration smoke samples in the field. Table 2 lists the 1995 experimentallyderived MDLs.

Many statistical tests and parameters are best suited to data with a normal (bell-shaped) distribution; however, much of the exposure sample data in this projectapproximated a geometric (lognormal) distribution. Whenever geometric distributionswere apparent but the statistical analyses required a normal distribution of data, theconcentration data were logarithmically transformed with equation (3) prior to theanalyses:

–LogX = Log(X) + 0.05x , (3)

where LogX = transformed concentration;

X = concentration of the exposure sample; and

–x = mean concentration of the data.

The results of the statistical analyses were converted back to the original units byusing equation (4):

–X = eLogX - 0.05 x , (4)

where e = natural logarithm base.

The addition of the constant (5 percent of the arithmetic mean) allows the use of zero-concentration data to calculate the geometric mean and is preferable to adding a con-stant of 1 to all observations—commonly done but inappropriate when many of theobservations are much smaller than 1 (Liedel and others 1977). Geometric standard

11

Method DetectionLimits

Data Analysis

deviations were calculated as the antilog of the standard deviation of the transformedvariables. The geometric standard deviation ranged between 1 for invariant data to 4for highly variable data. For some data analyses, non-normality remained even afterthe logarithmic transformation. In the case of small data sets, additional sampling wasneeded to determine the true data distribution. In other instances, the non-normalitysuggested additional grouping factors within the populations, which are noted below inthe “Results” and “Discussion.”

Correlations between pollutants were evaluated where pollutant pairs were measuredconcurrently with a sampling pack, including exposure samples from firefighters andfield replicates (which were essentially additional samples of smoke in or near thefire). Linear regression techniques were used to examine the interpollutant correla-tions. Data were excluded from this analysis if any of the following occurred:

• The two pollutants in question were not successfully sampled at the same placeand time (± 2 minutes).

• Bias was indicated for either of the pollutants based on field quality assurance (for example, any PM3.5 samples invalidated by visible nonrespirable particulatematter).

• Either pollutant concentration was below the method detection limit.

Because previous work at prescribed fires showed that benzene does not correlatewith the other pollutants in smoke exposure samples if the firefighters have beenworking with gasoline (Reinhardt and others 1997a), such samples were excludedfrom the data for benzene regressions. Also, to minimize error in the independentvariable (X) relative to Y, data for each regression were limited to instances where theX pollutant was at least twice the MDL. Finally, residuals were examined as a functionof possible confounding variables for each regression to ensure that the regressionmodels were unbiased (Neter and others 1983).

Time-weighted average (TWA) smoke exposures were calculated for each firefighterto assess shift-average and fireline-average exposure. Each TWA was calculated by:

12

Pollutant Correlations

Time-WeightedAverage Exposures

Table 3—Treatment codes for each period in the day comprising the time-weighted average––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––Code Treatment Summary––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––1 Cn=Cn Concentration of the entire period equals the sample

concentration obtained during the period, or portion thereof2 Cn=0 Concentration during the period is assumed to equal zero

(background)3 Cn=Cn+1 Concentration during the period is assumed to equal the

following sampled period4 Cn=Cn-1 Concentration during the period is assumed to equal the

previous sampled period5 Cn=TWACn-1,n+1 Concentration during the period is assumed to equal a

time-weighted average of the surrounding sampled periods––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

TWA =C1 3 T1 + C2 3 T2 +…+ Cn 3 Tn , (5)–––––––––––––––––––––––––––

T1

+ T2

+…+Tn

where Tn = duration of period n, and

Cn = pollutant concentration during period n.

The shift TWA included the paid hours from the start of the shift until its end. For project wildfires, lunch breaks were excluded from the TWA calculations unless theywere taken in smoky situations. Lunch or dinner breaks were assumed to be one-halfhour for calculating the TWA, even when work did not resume until somewhat later.

For initial attack wildfire crews, shift TWAs were defined differently. The CDF enginecrews technically worked a 96-hour on-call shift, followed by 72 hours off duty. Forthose crews, a work shift was defined for each day of sampling as the time elapsedbetween morning wake up and the end of fire-related duties in the evening, includingmeals. For Forest Service hotshot crews, each shift included the hours from the startof their scheduled shift until the shift ended (in one case, the shift “ended” when thecrew was dispatched to another fire out of the region). For these crews, unpaid lunchbreaks were excluded from the TWA calculations.

Along with the sampled periods comprising each TWA, unsampled periods occurred;these were treated consistently through a limited set of assumptions about smokeconcentrations during such periods. These assumptions were based on field observa-tions about the job task and smoke conditions for each period and expressed as oneof five codes defining the best assumption for that time period (table 3). The codes,period time, and relevant pollutant sample data for that firefighter were then used tocalculate an estimate of the concentration for the unsampled period.

Where data were missing for one or more pollutants in any sampled periods, the TWAfor that firefighter was calculated by using the interpollutant regressions and the sam-ple results for the other pollutants for that period. The best regressions for the missingpollutant, based on r 2 (coefficient of determination), were selected for the relevant firetype (initial attack or project wildfire). As an example, if a PM3.5 sample were invali-dated by a pump malfunction between 7:00 and 8:12 a.m., the CO data recorded dur-ing that time by the dosimeter on a firefighter would be used to estimate the missingPM3.5 concentration for that period for that firefighter. Regression results were aver-aged if equally efficient regressor pollutants were available, and the MDL was substi-tuted for any regression results that were below the MDL.

Samples from peak exposure situations were obtained in two ways: by using integrat-ed STEL sampling for all pollutants, and by identifying the peak CO exposure from thedata-logger results for CO in 1995. Only integrated samples with durations below 20minutes in peak exposure situations were considered valid as STEL samples. Therewere too few STEL samples from project wildfires to test whether they were differentfrom STEL samples at initial attack wildfires, so they were summarized separately.A Student’s T-test (Steel and Torrie 1980) showed no significant difference, however,between the log-transformed peak CO exposure data from 1995 at project wildfiresversus the data from initial attack wildfires. Because of this, they were combined foran overall peak CO exposure during wildland firefighting.

13

Peak Exposures

The job task categories we observed (and some common synonyms) were:

• Attack (direct attack): Intensive efforts to contain and extinguish the wildfire. Thisincluded hotline construction (direct line adjacent to burning or smoldering areas)and spotfire suppression, but not final mop-up. Laying hose to support a directattack was included in this task.

• Crew boss: Supervising a crew's field activities, keeping a lookout, and scoutingarea if necessary, usually performing tasks as well as managing them.

• Digging line: Indirect line construction and direct line construction in blackenedareas having few, if any, smoldering combustion sources. The use of hand tools wasincluded in this task, but sawyers and swampers were not included. Deployment offireline explosives was included in this task.

• Engine: This category of work encompassed all work near internal combustionengines, such as fire engine operators and water pump operators. Personnelassigned solely as vehicle drivers were neither sampled nor included in this catego-ry. For the initial attack wildfires, fire engine operators and captains that stayed withthe engine were included; at project wildfires, this category included only pumpoperators.

• Gridding: This activity involved patrolling systematically to find and extinguish spot-fires and hotspots. Some brief direct attack activities were included in this categorywhen hotspots were found.

• Holding (holding line, patrolling): Maintaining fire within fireline boundaries. Thisincluded using fire hoses and hand tools to maintain firelines and extinguish minorspot-fires along the fireline, periodic forays along and outside sections of fireline tocheck for spot-fires, and equipment and water supply maintenance work.

• Lighting: Use of a hand-held driptorch to ignite fuels during burnout operations.

• Mobile attack: This specific type of direct attack occurred only at initial attack wild-fires and was characterized by a firefighter working with a firehose in front of a slow-ly driven fire engine. This is a common tactic in herbaceous fuels and gentle terrain.

• Mop-up (mopping, dry-mopping, wet-mopping): This included using fire hoses,portable backpack pumps, and hand tools to extinguish smoldering woody debrisafter the main flaming phase of the fire had passed. Fire hose line maintenance,installation of branch lines, and equipment and water supply maintenance wereincluded. Patrolling for spot-fires in unburned areas within the firelines was includedin this task.

• Sawyer: Periodic operation of a chainsaw to fell trees and snags and cut updowned logs during line construction and mop-up.

• Swamper: Closely assisting the sawyer during line construction by maneuveringwoody material for cutting and clearing cut-up logs and branches from the path ofthe fireline.

14

Job Task

Geometric mean concentrations by job task were calculated for each pollutant. Datawere treated separately for initial attack wildfires and project wildfires. The log-trans-formed pollutant concentration data for each job task were tested for normality by theShapiro-Wilk procedure (Shapiro and Wilk 1965). Tukey's multiple comparison testwas used to examine whether the geometric mean exposures for each work task weresignificantly different (Steel and Torrie 1980).

With the exception of the 1994-95 fireline windspeed observations made with hand-held anemometers, site-specific environmental data from project wildfires were inade-quate. Conversely, we were able to obtain useful data from CDF for the initial attackwildfires in Redding, California. Weather, fuel moisture, and National Fire DangerRating System (NFDRS) predictions of fire behavior (Deeming and others 1977) werepaired by time with the smoke exposure data to assess relations among smoke expo-sure and windspeed, relative humidity, fuel moisture, predicted spread component,predicted burning index, and predicted ignition index. Windspeed data for these initialattack wildfires were obtained from an anemometer 6.1 meters in elevation at theRedding airport. The NFDRS predictions used a single fuel model for the Redding initial attack area, brush model F.

Scatterplots of smoke exposure versus environmental and fire behavior variables wereexamined for the initial attack wildfires. Based on the patterns observed, only wind-speed showed a definitive trend when plotted versus smoke exposure. Linear regres-sion was used to relate smoke exposure concentrations with windspeed for the combined data from initial attack wildfires and project wildfires.

Three different methods of measuring CO exposure were tested by collecting dataconcurrently via all three methods for randomly selected firefighters during the 1992through 1994 fire seasons:

• Infrared analysis of integrated gas samples (ICM 128, the reference method of COmeasurement for the project during 1992-94)

• Passive dosimeter tubes from Draeger® and Sensidyne® in 1992-94

• Passive electrochemical dosimeters in 1994 (using the Draeger model 190 data logger via OSHA method ID-209, which became the reference CO measurementmethod for the project in 1995)

Not enough CO results were within the measurement range of the dosimeter tubes tomake meaningful statistical comparisons with those data. The dosimeter tubes mayhave been adversely affected because the storage requirements of <25 °C could notbe met in the field. The CO exposure results from the remaining two methods werecompared for each sample period by using linear regression.

A visual estimate of the intensity of each firefighter’s exposure to smoke was routinelymade by the nearest observer during the work shift. These observations were usefulfor estimating smoke exposure for unsampled periods when calculating the overallTWA exposure for the firefighter’s work shifts. Ten different observers recorded smokeclassification data during the project. Smoke intensity during each obser vation was

15

EnvironmentalVariables

Dosimeters

Observer Estimates

classified as none (1), light (2), medium (3), heavy (4), and very heavy (5). If smokeconditions were changing rapidly, the recorded observation consisted of an entry foreach appropriate smoke class.

These data allowed us to test how well observers with little previous experience couldestimate smoke exposure. Because smoke conditions can change rapidly but ourobservers were unable to constantly watch the firefighters, we limited the data to sam-pled periods where the average time between smoke observations was no more than20 minutes. The smoke intensity observations were averaged for each sample period,and the concentration of each pollutant was plotted against the resulting visual smoke

16

Table 4—Index to wildfires comprising the study––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––Firenumber Fire type Fire name Date Fuel speciesa Location––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––20 Project County Line 8/6/92 MC Lowman, ID21 Project County Line #2 8/8/92 MC Lowman, ID22 Project Foothills Fire #1 8/24/92 G Idaho City, ID23 Project Foothills Fire #2 8/26/92 PP/MC Idaho City, ID24 Project Foothills Fire #3 8/27/92 PP/MC Idaho City, ID43 Initial attack Swaysey 8/17/93 CH/O/G Redding, CA44 Initial attack Akritch 8/18/93 G/O Redding, CA45 Initial attack Shawn/Paloma 8/22/93 O/G/CH Redding, CA46 Initial attack Cambridge 8/23/93 O/G Redding, CA47 Initial attack QH Ranch 8/25/93 O/P/G Redding, CA48 Initial attack Silverthorn 8/26/93 O/G Redding, CA49 Initial attack Squaw Grass 8/27/93 PP/MC Redding, CA50 Initial attack Misty Lane 8/29/93 O/G/CH Redding, CA55 Project Tyee Complex #1 8/1/94 MC/LP Chelan, WA56 Project Tyee Complex #2 8/3/94 MC/LP Chelan, WA57 Project Tyee Complex #3 8/5/94 MC/LP Chelan, WA58 Initial attack Virginia 8/13/94 G/O Redding, CA59 Initial attack Chip 8/16/94 G Redding, CA60 Initial attack Shasta View 8/17/94 G/O Redding, CA61 Project Libby Complex #1 8/25/94 MC Libby, MT62 Project Libby Complex #2 8/27/94 MC Libby, MT63 Project Ann #1 8/29/94 MC/LP Hamilton, MT64 Project Ann #2 8/30/94 MC/LP Hamilton, MT65 Project Covington 8/2/95 G/O/CH Joshua Tree, CA66 Initial attack Freeway 8/4/95 G/O Bear Divide, CA67 Project Verbenia #1 8/5/95 CH/O Cabazon, CA68 Project Verbenia #2 8/7/95 CH/O/PP Cabazon, CA69 Project Verbenia #3 8/8/95 CH/O/PP Cabazon, CA70 Initial attack Trask 8/12/95 CH/PP Monrovia, CA71 Project Helester 8/14/95 PP/O Tahoe, CA––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––aDominant overstory species listed in descending order of occurrence: CH—northern and southernCalifornia chaparral (mainly manzanita [Arctostaphylos spp.] and oak [Quercus spp.] in the north);G—western annual grasses such as cheatgrass (Bromus tectorum L.), medusahead ryegrass (Elymuscaput-medusae L.), and fescues (Festuca spp.); LP—lodgepole pine (Pinus contorta var. latifolia Engelm.);MC—mixed conifers (Douglas-fir, ponderosa pine, white fir [Abies concolor (Gord & Glend.) Lindl. exHildebr.], grand fir [Abies grandis (Dougl. ex D. Don) Lindl.], and western larch [Larix occidentalis Nutt.],predominantly on higher elevation sites of the interior West); O—oak (Quercus spp.);and PP—ponderosapine (Pinus ponderosa Dougl. ex Laws.).

estimate score for the period. After viewing scatterplots of the data, we evaluated linear regressions for each pollutant; measured pollutant concentration was used asthe dependent variable, and smoke exposure “class” was the independent variable.

Smoke exposure among firefighters was monitored during 30 days of wildfire suppres-sion between August 1992 and August 1995. Eighty-four firefighters were selected forexposure monitoring during 17 days at 8 separate project wildfires, and 45 firefighterswere monitored during 13 days of initial attack incidents. A total of 1,763 separatebreathing zone samples were collected to measure firefighter exposure to benzene,acrolein, formaldehyde, CO, CO2, respirable particulate matter, and total suspendedparticulate during wildfires. Table 4 lists the dates when firefighter exposure to smokewas monitored and the dominant vegetation that burned at each fire.

The results are organized by sections on the following:

• Data quality, to discuss the accuracy, precision and completeness of the exposuremeasurements as those parameters affect the conclusions

• Pollutant correlations, to show the strong links between exposures to different pollutants

• Exposure assessment at project wildfires, to summarize TWA smoke exposuresamong firefighters who have the potential for all-day smoke exposure at these multiday fires

• Exposure assessment at initial attack wildfires, to summarize the TWA smoke expo-sures among initial-attack firefighters who have no smoke exposure until they arecalled to respond to small wildfires

• Peak exposure assessment, to describe the highest smoke levels monitored at project wildfires and initial attack wildfires

• Factors influencing smoke exposure, to summarize the important determinants ofsmoke exposure

• Dosimeter performance assessment, to demonstrate the utility of this relatively newand simple procedure for measuring smoke exposure

• Observer estimates, to show the feasibility of visual estimates of smoke intensity formaking decisions about the safety of smoke exposure

Samples were analyzed by six methods for the seven pollutants. Some sampleresults were invalidated owing to technical problems in the field or laboratory. Table 5summarizes the number of samples attempted in the field, the successful field meas-urements, the number of results that have potential problems rendering them “quali -fied” (considered to be estimated values), and the percentage of completeness of thedata (how many were valid versus how many were attempted). Only valid data wereused to determine smoke exposure and correlations among parameters. Valid dataincluded individual data points that were “qualified.”

17

Results

Data Quality

A statistical analysis was performed on the overall set of QC sample data drawn fromthis project and a concurrent sampling effort using the same methods at prescribedburns (Reinhardt and others 1994). The larger set of QC data that resulted from com-bining the two data sets produced robust, statistically generated limits for accuracyand precision for the project team’s measurements (table 6). These performance-based QC limits were incorporated into the revised quality assurance project plan

18

Table 6—Accuracy and precision for smoke exposure assessment measurements––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Overall accuracy Overall precision(recovery) (relative standard deviation)

–––––––––––––––––––––––––– –––––––––––––––––––––––Statistically Statistically

Pollutant baseda Initial QA plan basedb Initial QA plan––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

PercentBenzene 75-119 90-110 24 30Acrolein 60-134 70-105 30 30Formaldehyde 65-130 80-110 43 30Carbon monoxidec 93-108d 90-110 31 15Carbon monoxidee 82-116 NA 19 NACarbon dioxide 98-103d 90-110 14 15Respirable particulate matter NA NA 32 20Total suspended particulate NA NA 20 NAf

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––NA = not applicable;analytical or sampling accuracy could not be measured.a Estimated by using liquid method spike recoveries and incorporated into quality assurance project plan(Radian Corporation 1996).b Estimated by using relative standard deviation of field replicates and incorporated into quality assuranceproject plan (Radian Corporation 1996).c Measured by intersociety committee method 128.d For analytical accuracy only;accuracy associated with sample collection and handling could not be measured.e Measured by OSHA method ID-209.f Analytical method implemented after the original quality assurance project plan was written.

Table 5—Summary of data collected for each pollutant––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Number of Number of Number ofsamples valid qualified

Pollutant Method attempted samplesa results Completeness

PercentBenzene NIOSH 1501 291 208 49 71Acrolein EPA TO-11 291 189 129 65Formaldehyde EPA TO-11 291 244 12 84Carbon monoxide Method 128b 259 227 30 78Carbon monoxide OSHA ID-209 49 37 0 100 Carbon dioxide Method 128b 259 201 19 78Total suspended

particulate NIOSH 0500 32 29 8 91Respirable

particulate NIOSH 0600 291 115 29 40––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––a Includes “qualified”sample results.b Intersociety committee standard methods.

(QAPP)(Radian Corporation 1996) and used to evaluate the quality of the wildfire datareported here. The accuracy and precision targets from the initial QAPP (RadianCorporation 1993) are included in table 6 for comparison. The qualified results indi-cated in table 5 are considered estimated values, which may not be within the statisti-cally based accuracy and precision listed in table 6. Pollutant identification was notaffected, only quantitation. In addition, all results within a factor of four of the associ-ated detection limit are estimated because measurement precision decreases at values near the detection limit.

Two sets of performance evaluation (PE) samples were analyzed during the project,as required by the QAPPs. Four additional sets were analyzed for acrolein andformaldehyde because of the difficulty of the method. Three of the six PE samples foracrolein and formaldehyde indicated a low bias, but corrective actions taken to identifythe source—including a review of all analytical procedures, method spikes, and certi-fied standard analyses—were inconclusive; therefore, our aldehyde results are to beconsidered potentially biased low by 50 percent (or a factor of two) and 66 percent (ora factor of three) for formaldehyde and acrolein, respectively. As well, all samples inwhich acrolein was not detected at the MDL were qualified to indicate the potential forthose results to be false negative. The potential of a low bias for formaldehyde andacrolein results and falsely negative acrolein results was considered in our interpreta-tion of the results.

Two laboratory systems audits and two field sampling audits were performed duringthe course of the project, as required in the QAPPs. The audit results indicated that,overall, the sampling procedures and analytical methods were producing data of suffi-cient quality for project use.

We found significant correlations among most of the pollutants. Only total suspendedparticulate was not well-correlated to other pollutants in smoke. Correlations with CO are especially important because of the ease of measuring CO in comparison to other pollutants. Correlations were expressed as linear regressions between pollu-tants. For example, figure 2 shows the regression between formaldehyde and COexposure levels at project wildfires. Each data point represents a sample pair whereboth pollutants were concurrently sampled from the same firefighter or field replicateat a fire.

All the initial attack wildfires were in or adjacent to urban areas, and project wildfireswere in more remote areas with little occurrence of urban pollutants. Strong correla-tions among the pollutants depended on their arising from a common source (thefires), yet the urban areas surely contributed to measured levels of several pollutants,such as CO. Because of this, the regression data were analyzed and applied separately for project wildfires (table 7) and initial attack wildfires (table 8) to avoidintroducing variability to the project wildfire regressions arising from urban source impacts affecting only the samples from initial attack wildfires. In the absence ofurban sources of pollutants, we do not expect the regressions to differ significantlybetween initial attack and project wildfires.

Pollutant correlations at project wildfires—Because the sample durations wereusually longer at project wildfires than at initial attack wildfires, more samples from theformer exceeded the MDLs. The greater number of available data pairs contributed tostronger regressions at project wildfires than those obtained at initial attack wildfires(see table 7).

19

Pollutant Correlations

20

Table 7—Interpollutant correlations in smoke samples from project wildfires–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Regression parameters (for the regression equation y=ax+b)–––––––––––––––––––––––––––––––––––––––––––––––––

Pollutants (y,x) na r 2 b a Std.error b Std.error–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––Formaldehydec

vs.carbon monoxidec 103 0.79 3.598 x 10 -3 1.816 x 10 -4 3.920 x 10 -3 2.911 x 10 -3

PM3.5d

vs.carbon monoxide 25 .79 4.980 x 10 -2 5.346 x 10 -3 8.135 x 10 -1 1.727 x 10 -1

Benzenec

vs.carbon monoxide 54 .91 1.089 x 10 -3 4.667 x 10 -5 3.399x 10 -3 1.085 x 10 -3

Acroleinc

vs.carbon monoxide 41 .68 4.200 x 10 -4 4.638 x 10 -5 3.260 x 10 -3 1.350 x 10 -3

PM3.5vs.formaldehyde 31 .68 18.19 x 100 2.317 x 100 1.705 x 10 -1 2.293 x 10 -1

Benzenevs.formaldehyde 62 .70 2.221 x 10-1 1.869 x 10 -2 6.698 x 10 -3 1.284 x 10 -3

Acroleinvs.formaldehyde 58 .81 1.841 x 10 -1 1.207 x 10 -2 -1.336 x 10-3 1.005 x 10 -3

Acrolein vs.PM3.5 14 .49 7.346 x 10 -3 2.184 x 10 -3 2.62 x 10 -4 5.999 x 10 -3

–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––a n = number of pairs of samples.b r2 = coefficient of determination.c Units in ppm.d Units in mg/m3.

Figure 2—Correlation between formaldehyde and CO in smoke samples at project wild-fires. The least squares regression between two pollutants is indicated by the solid line.For predicting formaldehyde levels at a given exposure to CO, 95 percent of the data up to 55 ppm CO will be within the error bands about regression, indicated by the dashed lines above and below the regression line.

21

Figure 3 shows the correlation between PM3.5 and CO exposure among firefighters at project wildfires. As shown in table 5, many data pairs were eliminated from theregression because nonrespirable particulates invalidated the PM3.5 samples.

Table 8—Interpollutant correlations for initial attack wildfires––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Regression parameters (for the regression equation y=ax+b)––––––––––––––––––––––––––––––––––––––––––––––––––

Pollutants (y,x) na r2b a Std.error b Std.error

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––Formaldehydec

vs.carbon monoxidec 39 0.59 5.638 x 10 -3 7.606 x 10-4 -1.463 x 10 -2 1.406 x 10 -2

PM 3.5d

vs.carbon monoxide 16 .46 9.909 x 10 -2 2.852 x 10 -2 6.883 x 10 -1 5.854 x 10 -1

Benzenec

vs.carbon monoxide 19 .44 1.275 x 10 -3 3.515 x 10 -4 1.199 x 10 -2 7.789 x 10 -3

Acroleinc

vs.carbon monoxide 20 .53 1.192 x 10 -3 2.626 x 10 -4 6.465 x 10 -3 5.812 x 10 -3

PM 3.5vs.formaldehyde 14 .79 23.35 x 100 3.473 x 100 2.889 x 10 -1 4.163 x 10 -1

Acroleinvs.formaldehyde 13 0.82 1.889 x 10 -1 2.654 x 10 -2 7.336 x 10 -3 4.414 x 10 -3

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––an = number of samples.br2 = coefficient of determination.cUnits of ppm.dUnits of mg/m3.

Figure 3—Correlation between respirable particulates and CO in smoke samples at project wildfires.

22

The correlation of benzene and CO is shown in figure 4. The large number of samplepairs strengthens the association shown between these pollutants.

Pollutant correlations at initial attack wildfires—The strength of the interpollutantregressions was generally lower at initial attack wildfires than at project wildfires. Thecorrelation between benzene and CO was especially low at these fires, which mightbe expected in an urbanized area because these contain many other sources of COand benzene that could impact the samples. The few data pairs that met the detec-tion limit criteria showed inadequate correlation to develop useful regressions betweenbenzene and formaldehyde and between acrolein and PM3.5 (see table 8).

Figure 5 shows the correlation between acrolein and CO samples obtained at initialattack wildfires.

Figure 4—Correlation between benzene and CO in smoke samples at project wildfires.

Figure 5—Correlation between acrolein and CO in smoke samples at initial attackwildfires.

23

Time-weighted average exposures were calculated for 84 firefighters at project wild-fires, including Forest Service type I (hotshot) crews, type II hand crews from multipleagencies, and three wildland engine crews. Average exposures to smoke for the timethat firefighters were at the fire (fire TWA) and for their entire work shift (shift TWA)were calculated from equation (5). The shift TWAs may include time spent in clean,ambient air while in transit between fire camp and fireline, but this usually makes up a relatively small proportion of the day at a project wildfire, whereas it can comprisemost of the day among initial attack forces.

Work shift durations averaged 13.9 hours for crews on project wildfires; of this, time onthe fireline averaged 10.4 hours. Figure 6 shows the distribution of fireline and workshift durations for the crews at project wildfires. Cumulative frequency distributionssummarizing the individual TWAs are presented in the following sections.

Acrolein—Acrolein exposure averaged 0.002 ppm on the fireline at project wildfiresand 0.001 ppm over the total work shift. The highest TWA acrolein exposure was0.016 ppm on the fireline and 0.015 ppm over a work shift. The geometric standard

ExposureAssessment forProject Wildfires

Figure 6—Distribution of work shift duration and time at project wildfires for firefighters.

Figure 7—Distribution of shift- and fireline-average acrolein exposure among firefighters at project wildfires.

24

deviation of the fireline and work shift TWA exposures indicate highly variable expo-sures at 3.6 and 4, respectively. Figure 7 shows the cumulative frequency distributionof acrolein exposures. Compare these values with the acrolein PEL and TLV of 0.1ppm.

Benzene—Exposure to benzene averaged 0.006 ppm on the fireline and 0.004 ppmover the work shifts. The highest TWA benzene exposure was 0.249 ppm over thework shift and 0.384 ppm on the fireline. The geometric standard deviation of theTWA exposures of the firefighters for both fire and shift was 3.6. Figure 8 shows thecumulative frequency distribution of benzene exposures for project wildfire crews.Compare these data with the benzene PEL of 1 ppm and the TLV of 0.5 ppm.

Figure 8—Distribution of shift- and fireline-average benzene exposure among firefight-ers at project wildfires.

Figure 9—Distribution of shift- and fireline-average CO2 exposure among firefighters at project wildfires.

Carbon dioxide—Carbon dioxide levels in the firefighters’ breathing zones averaged439 ppm over a work shift and 465 ppm at the project wildfires. The highest CO2 lev-els averaged 588 ppm and 668 ppm on the work shift and fireline, respectively. Thegeometric standard deviation of the shift and fireline TWAs indicated very consistentCO2 TWAs at 1.1 and 1.2, respectively. Some of the CO2 measured might have beenexhaled breath (CO2 is a product of normal human metabolism) rather than smokefrom the fires. Figure 9 shows the distributions of CO2 exposure at project wildfires.Compare these data with the PEL and TLV of 5,000 ppm.

Carbon monoxide—Exposure to CO among crews at project wildfires averaged 2.8ppm over the work shift and 4.0 ppm on the firelines. The maximum TWA exposure toCO was 30.5 ppm over the work shift and 38.8 ppm on the fireline. Geometric stan-dard deviations were 2.5 and 2.6 for fireline and work shift TWAs, respectively. Thedistributions of shift and fire TWA CO exposures among firefighters are shown in fig-ure 10. Compare these data with the PEL of 50 ppm and the TLV of 25 ppm.

25

Figure 10—Distribution of shift- and fireline-average CO exposure among firefighters at project wildfires.

Figure 11—Distribution of shift- and fireline-average formaldehyde exposure among firefighters at project wildfires.

Formaldehyde—Exposure to formaldehyde averaged 0.018 ppm on the fireline at the project wildfires and 0.013 ppm over the work shifts. The maximum exposure was0.093 ppm on the fire and 0.084 ppm over the work shift. The geometric standarddeviations for the TWAs were 2.3 and 2.4 for fireline and work shift exposures, respec-tively. Figure 11 shows the distributions of shift and fire TWA exposures to formalde-hyde at the project wildfires. Compare these data with the PEL of 0.75 ppm and theTLV of 0.3 ppm.

Respirable particulate—Exposure to PM3.5 among firefighters averaged 0.5 mil-ligram per cubic meter (mg/m3) over the work shifts and 0.72 mg/m3 on the fireline at project wildfires, with corresponding maximum exposures of 2.93 and 2.3 mg/m3.Geometric standard deviation of the work shift TWAs was 2.0 and 1.9 for the firelineTWAs. Figure 12 shows the cumulative frequency distributions for PM3.5 exposureamong the firefighters. Compare these data with the PEL of 5 mg/m3 and the TLV of3 mg/m3.

26

Figure 12—Distribution of shift- and fireline-average respirable particulate exposure among firefighters at project wildfires.

Figure 13—Distribution of shift- and fireline-average total suspended particulate exposure among firefighters at project wildfires.

Total suspended particulate—Exposure to total suspended particulate matter aver-aged 1.72 mg/m3 on the fireline at project wildfires and 1.47 mg/m3 over the workshift. The maximum TWA exposures to total suspended particulate were 4.17 mg/m3

on the fireline and 4.38 mg/m3 over a work shift. The geometric standard deviation ofthe TWAs was 1.8 for the fireline TWAs and 1.7 for the work shift TWAs. Few data areavailable because total suspended particulate exposure data were not collected untilthe 1995 fire season. Figure 13 summarizes the particulate matter exposure data.Compare these data with the PEL of 15 mg/m3.

Respiratory irritants—Exposure to the combination of respiratory irritants (acrolein,formaldehyde, and PM3.5) was calculated according to equation (1). Using OSHAPELs as the divisors in this equation, we calculated that exposure among firefightersaveraged 0.1 during the work shift and 0.1 on the fires. Maximum irritant exposureswere 0.6 for the work shift and 0.8 on the fireline. Geometric standard deviationswere 2.1 (work shift) and 2.0 (fireline). Figure 14 shows the distribution of the TWAexposure data, where the irritant mixture calculations are from the OSHA PELs. Thedata may be compared with the limit of 1.0 for the ratio, Em.

27

Figure 14—Distribution of shift- and fireline-average respiratory irritant exposure among firefighters at project wildfires (OSHA based).

Figure 15—Distribution of shift- and fireline-average respiratory irritant exposure among firefighters at project wildfires (ACGIH based).

28

Using the recommended ACGIH TLVs as the exposure limits in equation (1) results ina larger value for the irritant index than is calculated from the OSHA PELs. With theTLVs as the basis for calculating Em, exposure to respiratory irritants averaged 0.3(with a maximum of 1.4) on the firelines and 0.2 (maximum of 1.1) over the work shift.Figure 15 shows the resulting exposure distribution for the project wildfires. Again, thelimit of 1.0 for the ratio, Em, is the standard against which to evaluate the data.

Firefighters involved in initial attack suppression efforts included CDF wildland enginecrews and Forest Service hotshot crews. Average exposures to smoke were calculat-ed for the fire TWA and the shift TWA. The shift TWAs included time spent in clean,ambient air while performing other nonfire duties. This relatively large proportion ofthe day was the reason that the initial attack wildfire smoke exposure data were analyzed separately from the project wildfire data.

ExposureAssessment for InitialAttack Wildfires

Figure 16—Distribution of time on the fireline at initial attack wildfires and work shift durations for firefighters.

Figure 17—Distribution of shift- and fireline-average acrolein exposure among fire-fighters at initial attack wildfires.

On days when at least one initial attack event occurred, work shifts averaged 13.3hours for initial attack crews, and time on the fireline averaged 3.3 hours. Figure 16shows the distribution of time on the fireline at initial attack wildfires and work shiftduration for firefighters.

Acrolein—Acrolein exposure averaged 0.005 ppm (maximum of 0.037 ppm) duringinitial attack operations, and 0.001 ppm (maximum of 0.011) over a work shift. Thegeometric standard deviation for both TWA distributions was 4, indicating highly vari-able exposures. Figure 17 shows the cumulative frequency distributions of acroleinexposures. Compare these data with the PEL and TLV of 0.1 ppm.

29

Figure 18—Distribution of shift- and fireline-average benzene exposure among fire-fighters at initial attack wildfires.

Figure 19—Distribution of shift- and fireline-average CO2 exposure among firefighters at initial attack wildfires.

Benzene—Exposure to benzene averaged 0.014 ppm on the fireline and 0.003 overthe work shifts. The corresponding maximum exposures were 0.043 and 0.024 ppm.Geometric standard deviations were 3.2 for the fireline TWAs and 3.3 for the work shiftTWAs. Figure 18 shows the cumulative frequency distributions of benzene exposureamong initial attack crews. Compare these data with the PEL of 1.0 ppm and the TLVof 0.5 ppm.

Carbon dioxide—Carbon dioxide levels in the firefighters’ breathing zones averaged391 ppm over a work shift (maximum of 706 ppm) and 488 ppm (maximum of 742ppm) at the initial attack wildfires. The corresponding geometric standard deviationswere both 1.2. These results are limited to CDF engine crews because CO2 monitor-ing was not done in 1995. Figure 19 shows the distributions of CO2 exposure amongfirefighters at initial attack wildfires. These data can be compared with the PEL andTLV of 5,000 ppm.

30

Figure 20—Distribution of shift- and fireline-average exposure to CO among firefighters at initial attack wildfires.

Figure 21—Distribution of shift- and fireline-average exposure to formaldehyde among firefighters at initial attack wildfires.

Carbon monoxide—Exposure to CO averaged 1.6 ppm (maximum of 13.1 ppm) dur-ing the work shift among initial attack crews and 7.4 ppm (maximum of 28.2 ppm) atthe fires. Geometric standard deviations were 3.0 (work shift) and 2.2 (fireline). Thedistributions of shift and fire TWA CO exposures among firefighters are shown in fig-ure 20. These data can be compared with the PEL of 50 ppm and the TLV of 25 ppm.

Formaldehyde—Exposure to formaldehyde averaged 0.028 ppm (maximum of 0.092ppm) at the initial attack wildfires and 0.006 (maximum of 0.058) ppm over the workshifts. The corresponding geometric standard deviations were 3.0 and 3.1. Figure 21shows the distribution of shift and fire TWA exposures to formaldehyde at the initialattack wildfires. These data can be compared with the PEL of 0.75 ppm and the TLVof 0.3 ppm.

31

Figure 22—Distribution of shift- and fireline-average exposure to respirable particulates among firefighters at initial attack wildfires.

Figure 23—Distribution of shift- and fireline-average exposure to total suspended particulates among firefighters at initial attack wildfires.

Respirable particulate—Exposure to PM3.5 among firefighters averaged 0.022mg/m3 over the work shifts and 1.11 mg/m3 on the fireline at the initial attack wildfires.Maximum TWA exposures were 1.56 mg/m3 over the work shift and 2.46 mg/m3 onthe fireline. The corresponding geometric standard deviations were 2.5 and 1.6.Figure 22 shows the cumulative frequency distributions for PM3.5 exposure amongthe initial attack firefighters at wildfires. Compare these data with the PEL of 5 mg/m3

and the TLV of 3 mg/m3.

Total suspended particulate—Exposure to total suspended particulate matter averaged 5.32 mg/m3 (with a maximum of 8.64 mg/m3) at initial attack wildfires and1.39 mg/m3 (maximum of 1.81 mg/m3) over the work shift, based on only seven datapoints from Forest Service hotshot firefighters. Data on total suspended particulateexposure were not collected until the 1995 fire season, during which no CDF enginecrews were among those monitored. Geometric standard deviations were 1.4 for thefireline TWAs and 1.2 for the work shift TWAs. Figure 23 summarizes the exposuredata for the hotshot crews. Compare these data with the PEL of 15 mg/m3.

Respiratory irritants—Exposure to the combination of respiratory irritants (acrolein,formaldehyde, and PM3.5) was calculated according to equation (1). Using OSHAPELs as the divisors in this equation, we calculated that exposure among initial attackfirefighters at wildfires averaged 0.1 (maximum of 0.5) during the work shift and 0.4(maximum of 0.9) at the fires. The corresponding geometric standard deviations were 2.4 and 1.6. Figure 24 shows the exposure data at the initial attack wildfires.Compare these data with the OSHA PEL of 1.0 for the ratio, Em.

Using the recommended TLVs as the exposure limits in equation (1) resulted in ahigher irritant index. With the TLVs as the basis for calculating Em, exposure to respi-ratory irritants averaged 0.6 (maximum of 1.4) at the initial attack wildfires and 0.1(maximum of 0.8) over the work shift. Figure 25 shows the distribution of TWA irritantexposures among initial attack wildfire fighters. Compare these data with the TLV of1.0 for the ratio, Em.

32

Figure 24—Distribution of shift- and fireline-average respiratory irritants exposure among firefighters at initial attack wildfires (OSHA based).

33

Figure 25—Distribution of shift- and fireline-average exposure to respiratory irritants among firefighters at initial attack wildfires (ACGIH based).

Fire Carbon

number Fire type Time period monoxide

Ppm

59 Initial attack 16:49-17:03 8.1

59 Initial attack 17:01-17:15 3.1

60 Initial attack 16:53-17:07 23.1

60 Initial attack 16:53-17:07 18.5

60 Initial attack 16:55-17:09 18.2

66 Initial attack 14:06-14:20 39.5

66 Initial attack 14:12-14:26 48.6

66 Initial attack 14:24-14:38 47.5

66 Initial attack 14:23-14:37 39.3

70 Initial attack 17:55-18:09 47.1

70 Initial attack 17:36-17:50 25.6

70 Initial attack 17:29-17:43 11.1

70 Initial attack 17:39-17:53 14.9

55 Project 10:28-10:42 1.5

56 Project 09:51-10:05 31.1

56 Project 17:20-17:34 40.7

56 Project 14:38-14:52 15.8

57 Project 09:39-09:53 11.9

57 Project 10:27-10:41 22.8

61 Project 17:55-18:09 103.3

Fire Carbon

number Fire type Time period monoxide

Ppm

61 Project 17:22-17:36 77.1

62 Project 14:31-14:45 15.9

62 Project 14:31-14:45 18.4

62 Project 13:21-13:35 23.9

64 Project 14:16-14:30 15.5

64 Project 14:13-14:27 26.3

64 Project 14:12-14:26 30.3

65 Project 13:58-14:12 4.9

65 Project 14:00-14:14 5.4

65 Project 14:00-14:14 5.3

65 Project 14:11-14:25 1.4

67 Project 16:36-16:50 3.1

67 Project 16:22-16:36 2.7

67 Project 16:37-16:51 4.1

67 Project 16:24-16:38 2.3

68 Project 18:51-19:05 12.7

68 Project 17:44-17:58 19.5

68 Project 18:53-19:07 5.7

69 Project 12:43-12:57 10.1

69 Project 14:41-14:55 16.9

Table 9—Peak carbon monoxide exposures for 1994-95 wildfires

Carbon monoxide dosimeters facilitated measuring peak smoke exposures becausethey recorded data continuously; peak exposure events are easily extracted from thecontinuous record of exposure. Table 9 shows the peak CO exposure data from eachdosimeter record in 1994 and 1995. Data from 1992 and 1993 were not used be-cause of measurement bias, a problem resolved by improving the QA protocol for theinstruments in 1994 and 1995. Each observation is the highest 15-minute CO aver-age from the firefighter’s work shift. A few peak exposures included periods where theCO levels were briefly above 200 ppm. An overall lognormal mean of 13.7 ppm wascalculated for the 40 samples, which had a geometric standard deviation of 2.9.

34

Table 10—Peak exposure samples from project wildfires––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––Fire Carbon Formal-no. Duration Job task monoxide dehyde Acrolein Benzene PM3.5 Em––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Minutes -------------------- Ppm -------------------- Mg/m3

57 19 Attack NA 0.142 0.03 0.052 NA NA57 16 Attack 17.1 .077 0 .058 NA NA 61 20 Hold/mop 105.8 .282 .072 NA 5.5 1.8 64 10 Lighting NA .084 0 .077 1.8 .5 –––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––NA = not available.

Table 11—Peak exposure samples from initial attack wildfires––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––Fire Work Carbon Carbon Formal-no. Duration activity monoxide dioxide dehyde Acrolein Benzene PM3.5 Em

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––Minutes ---------------------------- Ppm ----------------------------- Mg/m3

43 16 Attack 14.3 640 0.16 0 0.038 3.17 0.9 43 13 Attack 16.3 532 .04 0 .036 1.26 .3 43 20 Attack 22.1 559 .136 .013 .022 NA NA 43 19 Attack 30.4 778 .181 .063 .024 NA NA 46 18 Attack 15.1 583 .038 .009 .019 NA NA 46 19 Attack NA NA .033 .007 NA 0.88 0.246 15 Attack 6.7 548 NA NA NA 1.14 NA 46 19 Attack 1.4 496 .031 .009 .022 .91 .247 16 Attack NA NA .153 .044 .041 3.05 147 13 Engine NA NA .16 .041 .041 2.93 147 17 Attack 38.7 1011 .168 .04 .079 5.4 1.3 50 15 Attack 21 759 .189 .037 .032 NA NA 50 13 Attack 42.2 836 .339 .066 .061 NA NA 50 18 Engine 34.6 888 .236 .051 .045 6.88 1.7 58 16 Attack 5.2 473 .073 0 .027 NA NA 58 20 Attack 8.4 863 .217 0 0 2.14 158 20 Attack 8.4 1265 .044 0 .022 NA NA 60 13 Mobile attack 4.7 450 .085 0 0 2.43 .6 60 15 Mobile attack 17.3 491 .032 0 .038 NA NA 60 16 Mobile attack 10.6 501 .071 0 .082 2.54 .5 60 16 Mobile attack 22.4 467 .05 0 .036 NA NA ––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––NA = not available.

Peak SmokeExposureAssessment

Project wildfires—Only a few peak exposure samples were collected with the sam-pling pumps at project wildfires. Many more transient peak exposure situations wereobserved but could not be sampled. Capturing data during peak exposure periods byusing a few observers and traditional sampling methods was much more difficult thanusing the data loggers because timing and logistics were critical to success. Table 10lists the results from peak exposure sampling during relatively smoky conditions at thegiven project fire. The equivalent irritant exposure index (Em) was calculated fromequation (1) and the recommended TLVs: 0.3 ppm formaldehyde, 0.3 ppm acrolein,and an excursion limit3 for PM3.5 of three times the TLV of 3.0 mg/m3 (AmericanConference of Governmental Industrial Hygienists 1996).

Initial attack wildfires—More peak exposure samples were collected at initial attackwildfires because the sampling plan called for short-term sample durations to capturebrief but intense exposures during the initial attack. Table 11 lists the peak exposuresamples obtained during the smokiest conditions observed at the initial attack wild-fires. The irritant exposure index (Em) was calculated as described above for projectwildfires.

Job task—Table 12 lists the geometric mean exposures during each major job task atproject wildfires. The duration of most samples was 1 to 2 hours. Because few sam-ples were obtained during some work activities, the means may not be truly represen-tative of exposure during that activity. Refer to the number of samples (n) in the tablesto evaluate the significance of the means.

35

3 An excursion limit is a short-term exposure limit for pollutants nothaving an otherwise-defined STEL or ceiling limit.

Table 12—Average smoke exposure by job task at project wildfires––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

TotalCarbon Carbon Formalde- suspended

monoxide dioxide Benzene hyde Acrolein PM3.5 particulate –––––––– ––––––– ––––––––– ––––––––– –––––––– ––––––––– ––––––––

Job task x–Ga nb x–G n x–G n x–G n x–G n x–G n x–G n

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––Ppm Ppm Ppm Ppm Ppm Mg/m3 Mg/m3

Direct attack 5.6 9 477 9 0.018 12 0.039 11 0.002 11 0.51 3 NA 0Crew boss 7.0 5 612 5 .010 4 .019 6 0 6 NA 0 NA 0Dig and attack 5.8 5 400 2 .000 2 .023 5 0 3 NA 0 12.00 4Digging line 3.9 21 472 21 .004 22 .022 21 .001 11 .99 7 NA 0Engine operator 13.7 2 586 2 .130 3 .079 3 .009 3 .42 1 NA 0Gridding 4.8 11 452 11 .004 12 .015 14 .002 9 .50 4 NA 0Hold and mop-up 45.1 7 750 7 .058 5 .098 9 .017 9 .71 5 NA 0Holding 3.2 15 465 10 .005 12 .015 17 0 12 3.75 10 1.43 4Lighting 8.7 6 569 6 .071 5 .094 8 .004 8 .34 6 NA 0Mop-up 4.4 52 477 46 .004 44 .022 54 0 33 1.84 24 4.33 9Sawyer 4.2 13 469 13 .021 9 .031 9 .001 3 .65 5 NA 0Swamper 3.7 3 469 3 .015 3 .027 3 .002 3 .67 1 6.44 1––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––NA = not available.ax–G = geometric mean.bn = number of samples.

Factors InfluencingSmoke Exposure

Figure 26 compares the distribution of CO exposures for workers during each job taskat project wildfires. The statistics were calculated on the logarithmically transformeddata, but these graphs show the data in original units. Figure 27 shows the distribu-tion of formaldehyde exposures by job task at project wildfires.

36

Figure 27—Distribution of formaldehyde exposure among firefighters by job task at proj-ect wildfires. The “box and whisker”plots use a box to show the data lying within the upper and lower quartiles (the middle 50 percent of the range). The solid square is the geometric mean of the samples, and the upper and lower vertical “whiskers”represent the data between the quartiles and the 95th and 5th percentiles. The relative length of the whiskers and the location of the mean depict the skew in the distribution (a sym-metrical arrangement of the box and whiskers indicates a normal distribution).

Figure 26—Distribution of CO exposure among firefighters by job task at project wildfires.The “box and whisker” plots use a box to show the data lying within the upper and lower quartiles (the middle 50 percent of the range). The solid square is the geometric mean of the samples, and the upper and lower vertical “whiskers” represent the data between the quartiles and the 95th and 5th percentiles. The relative length of the whiskers and the location of the mean depict the skew in the distribution (a symmetrical arrangement of the box and whiskers indicates a normal distribution).

Figure 28 compares exposure to benzene among different work activities at projectwildfires. At these project wildfires, the “engine” task was represented only by workerstending portable gas-powered water pumps.

Figure 29 compares exposure to PM3.5 by job task at project wildfires.

37

Figure 28—Distribution of benzene exposure among firefighters by job task at project wildfires. The “box and whisker”plots use a box to show the data lying within the upper and lower quartiles (the middle 50 percent of the range). The solid square is the geo-metric mean of the samples, and the upper and lower vertical “whiskers” represent the data between the quartiles and the 95th and 5th percentiles. The relative length of the whiskers and the location of the mean depict the skew in the distribution (a symmetrical arrangement of the box and whiskers indicates a normal distribution).

Figure 29—Distribution of respirable particulate exposure among firefighters by job task at project wildfires. The “box and whisker”plots use a box to show the data lying within the upper and lower quartiles (the middle 50 percent of the range). The solid square is the geometric mean of the samples, and the upper and lower vertical “whiskers” repre-sent the data between the quartiles and the 95th and 5th percentiles. The relativelength of the whiskers and the location of the mean depict the skew in the distribution (a symmetrical arrangement of the box and whiskers indicates a normal distribution).

Figure 30 shows total suspended particulate exposure during the work activitieswhere it was sampled. Fewer data are available for this comparison because TSPexposure data were not collected before 1995.

Table 13 summarizes the exposure to pollutants during the work activities at initialattack wildfires. The geometric mean and number of samples are listed for each activity at initial attack wildfires.

38

Table 13—Average smoke exposure by job task at initial attack wildfires––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––

Total Carbon Carbon Formal- suspended

monoxide dioxide Benzene dehyde Acrolein PM3.5 particulate––––––––– ––––––––– –––––––– –––––––– –––––––– –––––––– –––––––––

Job task x–Ga nb x–G n x–G n x–G n x–G n x–G n x–G n

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––– Ppm Ppm Ppm Ppm Ppm Ppm Ppm

Attack 11.6 22 653 21 0.021 22 0.069 21 0.004 22 1.74 12 12.19 1Crew Boss 9.6 2 NA 0 NA 0 .053 2 0 2 1.02 2 2.57 2Dig and attack 14.1 2 NA 0 NA 0 .045 2 0 2 NA 0 11.15 1Engine 8.7 6 547 6 .028 7 .041 8 0.003 7 1.77 6 NA 0Mobile attack 12.3 5 483 5 .020 5 .045 5 0 5 2.49 2 NA 0Mop-up 8.4 25 540 24 .007 29 .024 8 .003 26 1.05 19 1.00 1Sawyer 16.3 2 NA 0 .056 1 .059 2 0 3 NA 0 8.13 2Swamper 10.3 3 NA 0 NA 0 .035 3 0 2 .61 2 5.81 3

––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––NA = not available.ax–G = geometric mean.bn = number of samples.

Figure 30—Distribution of total suspended particulate exposure among firefighters byjob task at project wildfires. The “box and whisker” plots use a box to show the data lying within the upper and lower quartiles (the middle 50 percent of the range). The solid square is the geometric mean of the samples, and the upper and lower vertical “whiskers”represent the data between the quartiles and the 95th and 5th percentiles.The relative length of the whiskers and the location of the mean depict the skew in the distribution (a symmetrical arrangement of the box and whiskers indicates a normal distribution).

Figures 31 through 36 compare the CO, formaldehyde, benzene, respirable particles,acrolein, and total suspended particulate exposure among firefighters during differentwork activities at initial attack wildfires.

39

Figure 32—Distribution of formaldehyde exposure among firefighters by job task at initial attack wildfires. The “box and whisker” plots use a box to show the data lying within the upper and lower quartiles (the middle 50 percent of the range). The solid square is the geometric mean of the samples, and the upper and lower vertical “whiskers”represent the data between the quartiles and the 95th and 5th percentiles. The relative length of the whiskers and the location of the mean depict the skew in the distribution (a sym-metrical arrangement of the box and whiskers indicates a normal distribution).

Figure 31—Distribution of CO exposure among firefighters by job task at initial attackwildfires. The “box and whisker” plots use a box to show the data lying within the upper and lower quartiles (the middle 50 percent of the range). The solid square is the geo-metric mean of the samples, and the upper and lower vertical “whiskers” represent the data between the quartiles and the 95th and 5th percentiles. The relative length of the whiskers and the location of the mean depict the skew in the distribution (a symmetrical arrangement of the box and whiskers indicates a normal distribution).

40

Figure 34—Distribution of respirable particulate exposure among firefighters by job taskat initial attack wildfires. The “box and whisker” plots use a box to show the data lyingwithin the upper and lower quartiles (the middle 50 percent of the range). The solidsquare is the geometric mean of the samples, and the upper and lower vertical“whiskers” represent the data between the quartiles and the 95th and 5th percentiles.The relative length of the whiskers and the location of the mean depict the skew in thedistribution (a symmetrical arrangement of the box and whiskers indicates a normal distribution).

Figure 33—Distribution of benzene exposure among firefighters by job task at initial attack wildfires. The “box and whisker”plots use a box to show the data lying within the upper and lower quartiles (the middle 50 percent of the range). The solid square is the geometric mean of the samples, and the upper and lower vertical “whiskers”represent the data between the quartiles and the 95th and 5th percentiles. The relative length of the whiskers and the location of the mean depict the skew in the distribution (a sym-metrical arrangement of the box and whiskers indicates a normal distribution).

41

Figure 35—Distribution of acrolein exposure among firefighters by job task at initialattack wildfires. The “box and whisker”plots use a box to show the data lying within theupper and lower quartiles (the middle 50 percent of the range). The solid square is thegeometric mean of the samples, and the upper and lower vertical “whiskers”representthe data between the quartiles and the 95th and 5th percentiles. The relative length ofthe whiskers and the location of the mean depict the skew in the distribution (a symmet-rical arrangement of the box and whiskers indicates a normal distribution).

Figure 36—Distribution of total suspended particulate exposure among firefighters byjob task at initial attack wildfires. The “box and whisker” plots use a box to show thedata lying within the upper and lower quartiles (the middle 50 percent of the range).The solid square is the geometric mean of the samples, and the upper and lower verti-cal “whiskers”represent the data between the quartiles and the 95th and 5th per-centiles. The relative length of the whiskers and the location of the mean depict theskew in the distribution (a symmetrical arrangement of the box and whiskers indicatesa normal distribution).

Windspeed—Pollutant exposure levels were plotted against the corresponding ambi-ent windspeed at or near the fire to examine the relation between windspeed andsmoke exposure. Figure 37 shows the obser ved trend between ambient windspeedand CO exposures for firefighters engaged in various activities. Most of the data werefrominitial attack wildfires. Samples collected during mop-up did not show a trend relative to windspeed and are not plotted.

Other site factors—Carbon monoxide and formaldehyde exposure for the 11 initialattack wildfires in Redding, California, were plotted versus local data for relativehumidity, 1- and 10-hour fuel moisture, predicted rate of spread, burning index, andignition index. None of the plots indicated a trend between smoke exposure andthese site factors.

42

Figure 38—Correlation between carbon monoxide data from integrated bag samples and electronic dosimeters.

Figure 37—Correlation between CO exposure among firefighters and ambient windspeed.

Carbon monoxide measurements obtained at several wildfires in 1992-94 by usingICM 128 (Lodge 1989) were compared against concurrent data obtained with elec-tronic dosimeters by using OSHA method ID-209. The results indicate good linearity(the r2 is 0.97 for 38 sample pairs) but a negative bias in the dosimeter data relativeto the reference method 128 (fig. 38).

Figure 39 shows the response of the data loggers to a reference gas standard (aknown level of CO in air).Whenever possible, this QC check was performed beforeand after each work shift monitored. Each point on the x-axis represents a single datalogger at the given fire (from table 4); for example, two data loggers were used at fire55 and three at fire 56. Each point on the x-axis represents a single data logger at agiven fire (table 4). In most cases, the postsampling response (the empty triangles)was lower.

43

DosimeterPerformance

Figure 39—Field calibration check results for data logging CO dosimeters. Each point on the x-axis represents a single dosimeter at a given fire. The squares show the response of the dosimeter to the QC check gas before sampling a work shift, and the triangles mark the response at the end of a work shift.

Figure 40—Amount of CO standard gas measured by dosimeters in a smoke matrix.A = dosimeter A; B = dosimeter B;C = dosimeter C; and D = dosimeter D.

The 1995 QA protocol included testing the response of each dosimeter in the field toa known amount of CO mixed with a sample of smoke obtained from the fire. Aftersubsequent lab analysis of the CO level in the smoke sample (using ICM 128), wewere able to evaluate whether the dosimeter could accurately determine the amountof the “spiked” CO in the presence of the smoke matrix. We accomplished this test onseveral days of sampling; the data from the four dosimeters (identified as A through D)are summarized in figure 40.

The field observers’ classifications of firefighter smoke exposure were comparedagainst the actual concentration data (figs. 41-43). Linear regression was used toevaluate the ability of the observers to classify smoke exposure. The accuracy of theirestimates was measured by the fit of the regression line to the observed points (figs.41-43). The precision of their estimates is depicted by the envelope of the upper andlower 95-percent confidence intervals (for any given smoke exposure class estimateon the x-axis, 95 percent of the exposure samples would be within the band definedby the two confidence lines). Regressions were developed for CO, PM3.5, andformaldehyde. Insufficient data were available to develop a useful regression for totalsuspended particulate, and the data for acrolein were too widely scattered to makethis approach a useful tool.

Figure 41 shows how the actual PM3.5 exposure of the firefighters compared to the observers’ smoke classifications. A total of 46 PM3.5 samples were used in thesmoke classification regression analysis (two outliers were not used). For the smokeclasses ranging between none and medium-heavy (1 to 3.5), the r2 of the linearregression was 0.60. Equation (6) thus provides a very basic tool to estimate fire-fighter exposure to PM3.5:

PM3.5(mg / m3) = class 3 1.57(±0.19)-1.49(±0.37) , (6)

where class = average smoke intensity classification for the observation period.

44

Figure 41—Correlation between visual estimate and sampled respirable particulate exposure.

Observer Estimates

The standard error of the slope and intercept are given in parentheses after equationparameters. Figure 42 shows how the firefighters’ CO exposure associated with theobservers’ smoke intensity classifications. A total of 70 sample periods were used(two outliers are shown). The r2 for the regression was 0.37. Equation (7) providesusers with another basic tool to estimate CO exposure among wildland firefighters:

CO(ppm) = class 3 14.5(±2.3) - 14.2(±4.4) , (7)

Figure 43 shows how the firefighters’ exposure to formaldehyde differed compared to the observers’ smoke intensity ranking. A total of 68 sample periods were used for the regression analysis (omitting two outliers).The r2 for the regression was 0.39.Equation (8) summarizes this basic tool for estimating firefighter exposure toformaldehyde:

HCHO(ppm) = class 3 0.08(±0.01) - 0.09(±0.02) , (8)

45

Figure 42—Correlation between visual estimate and sampled CO exposure.

Figure 43—Correlation between visual estimate and sampled formaldehyde exposure.

Figures 44 through 48 show examples of our observers’ consensus rankings of smokeexposure—ranging from none to very heavy smoke exposure. Figure 44 shows a firecrew during a brief break in fireline construction at the Foothills wildfire in Idaho during1992, when no smoke was apparent to the observers. Equations (6) through (8) wereused to predict the average exposures for the no-smoke classification:

Carbon monoxide Formaldehyde Respirable particulate––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––0 ppm 0 ppm 0 mg/m3

Figure 45 shows a firefighting crew preparing to disembark down the fireline at theLibby Complex wildfires in Montana during 1994.The very weak inversion coupledwith little fire activity resulted in this light smoke exposure situation. Equations (6)through (8) were used to predict the average exposures for the light-smokeclassification:

46

Figure 44—Fire crew at the Foothills wildfire in Idaho, 1992.There was no smokeapparent to the observers.

Figure 45—Fire crew during the Libby Complex wildfires in Montana, 1994.The observ-ers classified this as a light smoke exposure situation.

Carbon monoxide Formaldehyde Respirable particulate––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––15 ppm 0.1 ppm 2 mg/m3

Figure 46 shows a crew holding line and mopping up during a late-afternoon burnoutoperation at the Libby Complex wildfires in Montana during 1994. The smoke expo-sure was classified as medium intensity by the observers. Equations (6) through (8)were used to predict the average exposures for the medium-smoke classification:

Carbon monoxide Formaldehyde Respirable particulate––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––30 ppm 0.15 ppm 3 mg/m3

47

Figure 46—Firefighting crew holding line during a late afternoon burnout operation in the Libby Complex wildfires in Montana, 1994. The observers classified this as a medium smoke exposure situation.

Figure 47—Firefighter holding a fireline. The observers classified this as a heavy smokeexposure situation.

Figure 47 shows a firefighter holding a fireline in heavy smoke. The smoke concentra-tion continued to increase during the observation period. Equations (6) through (8)were used to predict the average exposures for this smoke classification, based ondata from this study. Because the heavy-smoke classification extrapolates beyond thesample data of this study (none of our samples were from a period classified as“heavy smoke” for the entire period), they may be inaccurate:

Carbon monoxide Formaldehyde Respirable particulate––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––44 ppm 0.2 ppm 5 mg/m3

Finally, figure 48 shows the same firefighter as shown in figure 47 but in a very heavysmoke situation.The smoke exposure was relatively long and the firefighter sufferedfrom extreme nausea. Air sampling data from this study were used to predict theapproximate average exposures for this smoke classification from equations (6)through (8). Again, because these predictions extrapolate beyond the sample data ofthis study, they may be inaccurate:

Carbon monoxide Formaldehyde Respirable particulate––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––58 ppm 0.3 ppm 6 mg/m3

Most of the time, firefighters are not overexposed to smoke because they can remainupwind of fires during suppression. Work occasionally occurs, however, in thickersmoke, and these exposures can easily exceed STELs recommended by ACGIH.When peak exposures repeatedly occur, or are combined with extended work in mod-erate smoke, shift-average smoke exposure also can exceed recommended exposurelimits for the work shift.

Figures 10 and 15 show that up to 5 percent of the firefighters’ shift-average exposures to CO and the sum of respiratory irritants (acrolein, formaldehyde, and respirable particulate matter) exceeded ACGIH TLVs, and up to 10 percent of the

48

Figure 48—The same firefighter as in figure 47 holding a fireline. The observers classi-fied this as a very heavy smoke situation.

Discussion

firefighters exceeded these TLVs while working on a fireline. The CO and respiratoryirritant exposures summarized in figures 10 and 14 seem to be within the currentOSHA PELs, unless the allowed CO PELs are adjusted downward to account for thelong work shifts at wildfires. This adjustment results in a lower PEL, one exceeded by some of the measured exposures. The adjustment factor is obtained by dividing 8 hours by the actual shift length (14 hours, based on our data for project wildfires).This results in an adjusted PEL of 29 ppm CO, an enforceable limit that was exceededby about 3 percent of firefighters at project wildfires. Finally, considering the study’sQA results, up to 2 percent of the shift-average respiratory irritant exposures alsocould exceed the PELs because the imprecision of the irritant exposure measurementranges between 30 and 40 percent (table 6), and the aldehyde measurements maybe biased low by a factor of two to three.

The TSP data we began to obtain in 1995 were from conditions representing only thelower to middle range of smoke exposures, based on concomitant levels of the othercharacteristic pollutants in smoke. The TSP data indicated compliance with the OSHAPEL, but the true range of TSP exposure can be further assessed by additional moni-toring to ensure that TSP is not an overlooked problem. Benzene was not an inhala-tion hazard among firefighters, even among those occasionally working with gasoline.

Most of the firefighters’TWA pollutant exposures were brought into compliance bylong periods in the day without significant smoke exposure, but work on the firelinecan be in high-exposure conditions—especially during direct attack, at initial attackwildfires, and while holding a fireline during burnout operations. As shown in tables10 and 11 and figure 26, peak exposures to CO and respiratory irritants are likely toexceed recommended STELs in these situations. The probability of overexposure tosmoke is enhanced when firefighters are on the flanks or downwind of a fire in highambient winds, or when inversion conditions prevent smoke dispersal by trappingemissions from a large fire within a valley.

The exposure data we report are similar to results of other researchers. Results ofcarboxyhemoglobin (COHb) monitoring at 11 western wildfires in 1974-76 showed that9.4 percent of firefighters had postfire COHb levels above 5 percent (Jackson andTietz 1979). This value ranged from 0 to 100 percent of the firefighters tested at eachfire, highlighting the variability of smoke exposures. These COHb results would corre-spond to somewhat higher CO exposures than we measured, which could be due toinherently smokier situations at the fires monitored by Jackson and Tietz, and theyalso may be explained if there was, as some suggest, less concern about the adverseeffects of smoke in that era. More recently, Reh and Deitchman (1992) obtainedbreathing zone samples of CO during the 1988 project wildfires in YellowstoneNational Park that ranged between 3.6 and 7.8 ppm during mop-up and between 1.9and 3.9 ppm during a day of fireline construction—ranges consistent with our results.They also obtained area samples for total particulate matter, which were mostly below1.2 mg/m3, although one 5-hour sample was 15.9 mg/m3, and a 4-hour sample was47.6 mg/m3. Our results, again, are consistent with those and highlight the potentialfor occasionally high exposure levels. A few area samples for aldehydes detectedformaldehyde levels averaging between 0.02 and 0.03 ppm; benzene samples were all below 0.03 ppm. Their results are very consistent with our project wildfire results,especially for activities with lower smoke exposure potential and no gasoline exposure.

49

Finally, a summary of results from exposure monitoring in northern California between1986 and 1989 found that 46 samples of firefighter exposure to CO averaged 14.4ppm during fireline and mop-up activities at prescribed burns and wildfires; range,1.4 to 38 ppm (Materna and others 1992). Their mean CO exposure is higher thanour data, but the range is consistent. They also found that 22 samples of respirableparticulate exposure averaged 1.75 mg/m3 (range, 0.327-5.14 mg/m3), and total par-ticulate averaged 9.46 mg/m3, (range, 2.7 to 37.4 mg/m3). Again, the ranges of thesevalues are consistent with our results and point out the possibility of higher exposures.

Our data show that the respiratory irritants, formaldehyde, acrolein, and respirableparticulate, are well-correlated with CO, and that electronic dosimeter technology canbe an effective basis for a routine smoke exposure monitoring program on a broadscale. Our experience also was that obser vers can roughly estimate smoke exposurewith sufficient precision to determine whether administrative controls or respiratoryprotection should be used to reduce smoke exposures.

Exposure of firefighters to the key respiratory irritants in smoke can be predicted frommeasurements of CO, at least in the Western United States. Carbon monoxide levelsin smoke at project wildfires were strongly correlated to concentrations of formalde-hyde, PM3.5, and acrolein, as indicated by the r2 values (range, 0.44 to 0.91; tables 7 and 8). This opens up a cost-effective way to routinely assess exposure to manypollutants in smoke that are difficult and expensive to measure. Benzene and COexposure are especially well correlated when there is no exposure to other benzenesources such as gasoline or engine exhaust.

Although few samples were obtained, TSP was not significantly correlated to PM3.5either at project wildfires or during initial attack. If further assessment of TSP expo-sure indicates a significant hazard in more intense smoke exposure situations, it will be a key pollutant to routinely monitor because of the lack of correlation to CO.Emission measurements collected in the plume of prescribed burns (Sandberg andothers 1989) have shown strong correlations between concentrations of TSP and fineparticles (PM2.5); thus the correlation of TSP and PM3.5 in smoke also may be rela-tively strong. The lack of a similar correlation in our data is clearly due to entrainedsoil dust from firefighters’ activities dominating the samples in the absence of high levels of smoke.

To provide the best fit to the most data in each regression, we excluded a few outliersfrom the final regressions, based on poor fit compared to the overwhelming balance ofthe data; this is rightly of concern to statisticians. However, in so much as any of thesamples could be inaccurate because of indeterminate errors in the rugged field envi-ronment (such as unobserved temporary kinks in air sampling lines), we are confidentthat excluding the few outliers (from none to five excluded from up to 103 samplepairs, depending on the correlation) did not bias the accuracy and provided apprecia-bly more precise fits to the data. We used the Cook’s D and DFFITS statistics to iden-tify observations for possible exclusion (SAS Institute 1989).

The regressions with CO provide a useful quantitative tool to estimate Em, proven overa range of 1.5 times the PEL for Em (the mixture of acrolein, formaldehyde, andPM3.5). Fire managers may want to consider using the project wildfire regressions forthis purpose rather than the initial attack data from urban interface fires, because the

50

Pollutant Correlations

project wildfire regressions (1) are based on a larger number of samples, (2) cover agreater geographic range, and (3) provide greater precision than the initial attackregressions, which were likely to have been compromised by urban sources of pollu-tants. Fire managers may want to consider using these interpollutant correlations tocost-effectively monitor smoke exposure, but we caution that data from other regionsare necessary to determine whether the correlations are consistent over a broadergeographic scale.

Project wildfires—The interpollutant regressions at project wildfires were surprisinglystrong. The correlations between CO and each of the respiratory irritants can be usedto effectively estimate total irritant exposure from simple CO measurements in thefield. To do this, equation (9) may be conveniently programmed into a calculator ordistributed to crew supervisors as a nomogram:

Em =([CO] 3 3.598 3 10-3 + 4 x 10 -3)

+([CO] 3 4.98 3 10-2 + 0.8)

+([CO] 3 4.2 3 10-4 + 3 3 10-3) , (9)_________________________ _____________________ _______________________

(formaldehyde exposure limit) (PM 3.5 exposure limit) (acrolein exposure limit)

where Em is the total irritant exposure, and the appropriate exposure limit for each pollutant is chosen from table 1. A numerical example using the OSHA PELs is givenbelow. At a firefighter’s TWA CO concentration of 43 ppm over 8 hours, the estimatedtotal irritant exposure at a project wildfire would be:

Em =(43 3 3.598 3 10-3 + 4 3 10-3)

+(43 3 4.98 3 10-2 + 0.8)

+(43 3 4.2 3 10-4 + 3 3 10-3)________________________ ___________________ _____________________

(0.75) (5) (0.1)

Em = 1.01.

This example assumes that the PELs are adequate exposure limits for firefighters, andshows that a CO exposure for 8 hours that is below the current PEL of 50 ppm resultsin an irritant exposure just exceeding the OSHA PEL of 1.0. Here is another examplewith the recommended ACGIH TLVs as the evaluation criteria and an assumed TWACO exposure of 21 ppm over 8 hours:

Em =(21 3 3.598 3 10-3 + 4 3 10-3)

+(21 3 4.98 3 10-2 + 0.8)

+(21 3 4.2 3 10-4 + 3 3 10-3)________________________ ___________________ _____________________

(0.3) (3) (0.1)

Em = 1.00.

When using the ACGIH TLVs as the evaluation criteria, CO exposure must be main-tained below 21 ppm to keep the irritant exposures within acceptable limits. Whenselecting a CO exposure to use as an action level to prevent overexposure to irritants,keep in mind that the equations above are for the best-fit linear regression; therefore,some irritant exposure samples would be above as well as below the regression line.Also, if the formaldehyde and acrolein results we obtained are actually biased low, asindicated by some of our QA data, the actual irritant sum will be higher as the regres-sions would underpredict the true formaldehyde and acrolein levels.

51

Initial attack wildfires—Pollutant correlations at initial attack wildfires were not asstrong as at project wildfires. Compare the regressions in tables 7 and 8. Three rea-sons could explain the differences between the two sets of regressions: (1) more sam-ples were obtained at project wildfires, which increases the r2 for a regression if otherfactors are held constant; (2) the range of sample concentrations at project wildfireswas broader than at initial attack wildfires for all regression pairs except formaldehydeand acrolein, which also tends to increase an r2; and (3) the project wildfires weremuch further removed from urban sources of air pollution, so the relations among thepollutants in smoke were not obscured by contributions from urban sources. This latter possibility exists with the data from initial attack wildfires. For example, at lowersmoke concentrations, a small amount of background, urban CO (3 or 4 ppm) is a rel-atively large proportion of the total CO in the sample, which obscures true correlationsamong the pollutants in biomass smoke. If further data without these limitations wereobtained, the resulting regressions from initial attack wildfires and project wildfiresmight be equivalent.

Averaged over a work shift, smoke exposure at these wildfires was usually below theOSHA PELs, but a small percentage of work shifts exceeded the PEL for CO afteradjustment for the longer work shifts, under the OSHA approach. More would likelyexceed the 5 percent COHb criterion if the CFK equation were applied. The ACGIHrecommended TLVs were occasionally exceeded outright for CO, as well as for thesum of the respiratory irritants (acrolein, formaldehyde, and PM3.5). All benzeneexposures were well below the OSHA PEL, and only one exposure by a firefighter tobenzene exceeded the benzene TLV while that person was on the fireline. The peaksmoke exposure samples from project wildfires and initial attack wildfires were similarin concentration: all were below the OSHA STELs, but roughly half of the samplesexceeded an Em based on the TLV STELs for respiratory irritants. We can concludethat smoke exposure is more often a problem for brief periods within a day than overan entire work shift. In most cases where the shift-average exposure limits areexceeded, the exposures could be brought into compliance by focusing control efforts to reduce the peak exposures within the work shift.

Firefighters work long days at project wildfires, and their extended work shifts areinconsistent with the 40-hour workweek on which the PELs are based. The firefighterswe observed spent, on average, over 10 hours on the fireline within their 14-hour workshifts (fig. 6). Firefighters routinely labor through extended work shifts at project wild-fires and seldom receive optimum work and rest cycles. Such conditions warrantadjusting the exposure limit to achieve adequate protection against chemical hazards,especially CO.

Respiratory irritants—Although each respiratory irritant was individually below itsrespective PEL and TLV, the collective irritation of the eyes and respiratory system bythese smoke components was significant (figs. 14 and 15). Compared with the PELs,the highest TWA fireline and work shift exposures to irritants were about 80 percentand 65 percent of the PEL (Em=1.0), respectively. Using the ACGIH-recommendedTLVs as the evaluation criteria, about 8 percent of the firefighters’ fireline TWAs and 3 percent of the work shift TWAs exceeded the respiratory irritant TLV.

52

ExposureAssessment

ExposureAssessment forProject Wildfires

As discussed under “Data Quality,” above, some performance evaluation audit samplesindicated that the aldehyde measurements may be biased low. Review of our data,calculations, and procedures could not explain the discrepancy between low perform-ance-evaluation audit results for the aldehydes and the adequate results of field matrixspike recoveries (accuracy data from table 6). If a bias was present it was consistent,because the interpollutant correlations showed consistent ratios between the aldehy-des and the other pollutants in smoke that were accurately measured. If this biasexisted, it is probable that a small percentage of the exposures would have exceededthe OSHA irritant PEL at project wildfires, and a larger percentage would haveexceeded the ACGIH irritant TLV. To evaluate this, we corrected for the effect of apotential bias in our measurements of formaldehyde and acrolein and produced theresults illustrated in figure 49, which may be compared with figure 14 (in which thedata are assumed to have no bias).

Simultaneous measurements of the three respiratory irritants were successful onlyduring one of three peak exposure samples at project wildfires (table 10). This samplewas obtained from a firefighter holding fireline and mopping-up during a burnout. Thesample data indicated a peak exposure about twice the ACGIH guidelines but just 75percent of a PEL-based STEL for Em. We note that exposure during the smokiest 15minutes of this 20-minute sample (consistent with the STEL definition) may have beengreater than the sample obtained. In spite of our incomplete data, we can state thatpeak irritant exposures at project wildfires can reach levels that are very irritating toworkers. To summarize exposure to irritants at project wildfires, the data we haveindicated that the highest exposures are probably compliant with the current PELs,but they do not meet the guidelines established by ACGIH to prevent adverse effects,primarily irritation of the eyes and mucous membranes and temporary depression ofrespiratory system functions.

Benzene—Only one TWA benzene exposure for a firefighter exceeded the ACGIHTLV of 0.3 ppm at project wildfires. This exposure was estimated at 0.384 ppm over 9 hours on the fireline (fig. 8). No benzene exposures exceeded the OSHA PEL. Thefirefighter with the highest benzene exposure was igniting a burnout with a drip torch

53

Figure 49—Shift- and fireline-average respiratory irritant exposure at project wildfires,assuming biased aldehyde data.

in the morning and alternately carrying fuel, operating a chainsaw, and mopping-upduring the afternoon. His average benzene exposure over the entire shift was 0.249ppm, but his benzene exposure was 0.824 ppm during one 90-minute period of chain-saw operation. The firefighter with the next highest benzene exposure (0.104 ppm forthe shift, 0.133 ppm on the fireline) was tending a portable water pump during most ofthe work shift, well removed from the fire smoke. The third highest benzene exposure(0.066 ppm during the shift, 0.071 ppm on the fireline over 14 hours) was anothersawyer, who operated or carried a chainsaw most of the day. These results indicatedthat sawyers and other engine operators are most likely to have the highest benzeneexposures among fire crews. Workers involved in large refueling operations, such asthose preparing drip torches or helitorches, could have more significant benzene exposures. Such workers were not sampled in this study, but based on our results,fire management agencies may wish to target them for future benzene exposureassessment. We can conclude from these data that the benzene exposure of a typicalfirefighter working with hand tools is unlikely to exceed the current PELs or TLVs.

The two peak exposure samples from project wildfires that had successful benzeneresults were both around 0.055 ppm (table 10), well below either ACGIH or OSHASTELs. Those peak exposure samples were identified by their match to a peak smokeexposure situation, not by peak benzene exposure caused by gasoline vapors. Thelow benzene levels in the peak smoke samples reinforced our conclusion that neitherthe OSHA or ACGIH STELs for benzene are likely to be exceeded in the absence ofgasoline vapors.

Carbon dioxide—Carbon dioxide was not sampled in 1995, but the data from 1992-94 show that levels of CO2 measured in the breathing zone were relatively low, muchless than the OSHA PEL of 5,000 ppm (fig. 9). Some of the CO2 measured may befrom normal metabolic waste (exhaled breath) rather than forest fires. We did notexpect CO2 to be a health concern among firefighters, and the data bear this out.The data may be useful to future investigators interested in the effects of low-levelCO2 exposure. To that end, we note that background CO2 levels were not measured,and a background concentration of 346 ppm was assumed for estimating TWA CO2exposures.

Carbon monoxide—Carbon monoxide exposures among firefighters have a patternsimilar to respiratory irritants. The distribution of exposures is skewed, with most atlow concentration and a small percentage of higher exposures (fig. 10). No measuredexposure exceeded the current unadjusted OSHA PEL (50 ppm), one exceeded theformer OSHA PEL (35 ppm) while the firefighter was on the fireline, and less than 5 percent of the CO exposures exceeded the ACGIH TLV of 25 ppm.The OSHArequires that the OSHA PEL be adjusted to account for the potential COHb buildupduring a long work shift. One simple method is to use equation (2). For a 14-hourwork shift, this results in an adjusted CO PEL of 29 ppm, a limit exceeded for about 3 percent of our firefighters. The best method of adjusting the CO PEL uses the CFKequation to arrive at the equivalent limit to prevent a COHb level above 5 percent.The CFK equation gives a much lower exposure limit for a long work shift, especiallyin high-altitude conditions; an exposure limit of around 20 ppm would typically be calculated. In such a scenario, about 10 percent of the observed CO exposures would exceed the adjusted CO exposure limit.

54

Among the peak exposure data, the highest CO sample was just over 100 ppm, withmost samples in the range of 10 to 50 ppm. The CO values on the data logger for themore highly exposed firefighters were occasionally above 200 ppm for brief periods;if maintained for 5 minutes, they would exceed the former OSHA ceiling limit for CO.When peak CO exposures occur while firefighters are exerting maximum effort, COHb levels in the bloodstream can rise rapidly, on the order of 1 percent per minute.Because CO alone has no odor or warning properties, it is probably the best pollutantto be monitored in any routine exposure monitoring program. Carbon monoxide iseasy to monitor in high-exposure situations with electronic dosimeters; these devicescan sound alarms when CO levels exceed preset limits, and the strong correlationswith the respiratory irritants can be used to manage exposures to respiratory irritants.

Total suspended particulate—The firefighters’TWA exposures to TSP were all lessthan half of the exposure limits for nuisance dust (fig. 13). Based on these data, totalsuspended particulate exposures are not significant. The TSP data were collected,however, for only a small subset of the monitored firefighters (at 3 days of wildfire suppression in southern California), a set that was obtained in relatively smoke-freeconditions, based on the other pollutants measured among the same firefighters.The smoke exposure data obtained from 1992 to 1994 included much higher levels ofPM3.5 and CO (figs. 10 and 12) than were found in the data from 1995. For example,among those firefighters for whom both TSP and CO exposures were measured in1995, the highest TWA CO exposure was only 2.3 ppm on the fireline and 1.8 ppmover the work shift. The corresponding PM3.5 levels were all below 1 mg/m3. Thus,the TSP values that exceeded the PM3.5 levels (ranging up to 4.5 mg/m3) were associated with large particles of disturbed ash and soil dust rather than smoke.

We believe it is reasonable to assume that higher levels of smoke will be associatedwith higher levels of TSP. Fire emissions research has sampled concentrations of fine particles (PM2.5) and TSP in smoke plumes above operational prescribed burns(Sandberg and others 1989). The PM2.5 comprised about 56 percent of the TSP.Because the PM2.5 and PM3.5 portions of TSP in smoke are of similar size (and thusconcentration), firefighters exposed to 2.5 mg/m3 of PM3.5 are expected to have TSPexposures from smoke on the order of 5 mg/m3. Similarly, the correlation between COand TSP found by Sandberg and others (1989) suggests that the higher TSP expo-sures from smoke are about 6 mg/m3. When these estimates of the TSP contributedfrom smoke were added to the observed TSP levels from 1995 (which ranged up to4.5 mg/m3), an estimated exposure to TSP in a smoky environment was about 10mg/m3, which is two-thirds the PEL for TSP.

The results suggested that further monitoring of TSP exposure is warranted to establish whether compliance with the PEL is a problem during work shifts whensmoke levels are higher. The health significance of exposure to nonrespirable dustswithout fibrogenic or chemical irritant properties is limited to irritation so far as weknow; but the dust samples have not been analyzed for crystalline silica, and otherinvestigators have found significant levels of crystalline silica in similar samples fromfirefighters in northern California (Materna and others 1992).

We believe that our data are representative of average conditions, especially for COand PM3.5. The aldehyde data may be biased low, but we are not convinced of thisbecause of the strength of the interpollutant correlations. We emphasize that the

55

distribution of smoke exposures is skewed (there are many more low-exposure hoursthan high-exposure hours in a firefighter’s career); therefore, we are unlikely to havemeasured the highest exposures that occur. To do so would take a substantial monitoring effort because the lag time to arrive on-site is separate from the incidentresponse. Further research-scale collection of smoke exposure data, even if it cap-tures the occasional very high exposure, is unlikely to change smoke managementdirection for minimizing exposure levels.

We point out that our data are not without potential internal biases. Multiple firefight-ers were sampled each day, but the exposures we measured were not completelyindependent because (1) the firefighters were mostly from the same crew and workingthe same area of the wildfire, and (2) we expect that smoke exposures are more similar within a crew working in one area of a fire than between crews in differentareas. Because of this, we believe that more broadly based sampling is better suitedto define the distribution and upper bound of smoke exposures.

Although we monitored smoke exposure during 17 days at project wildfires, the datafrom project wildfires were obtained at only eight separate fires. Most of the projectwildfire sampling occurred on sequential days at the fires. It seems apparent from ourobservations that the urgency of firefighting efforts generally declines to a more meas-ured pace as a fire progresses from initial attack through days and weeks of contain-ment efforts and, finally, control and mop-up. As a fire progresses from a flamingphase to a smoldering phase, the smoke production rate declines as well. Becausemost of our sampling data from project wildfires were in the mid to latter phases of thefires, the exposure distributions we found may be biased low if higher exposures occurearlier in the fires. Future monitoring programs should strive to obtain enough dataover the course of a fire to assess the time series trend of smoke exposure over multiple days.

Initial attack crews had lower TWA smoke exposures than crews on project wildfires(figs. 6 and 16). Firefighters have about the same working shift length at initial attackand project wildfires, but the initial attack forces spend much less time on the fire-line–an average of only 3.3 hours at initial attack wildfires compared with 10.4 hours at project wildfires. Because the amount of time spent at the wildfire scene is so muchlower, shift-average exposure limits are less likely to be exceeded among initial attackcrews, at least when only one or two small fires occur per day. The data showed thatoverexposure to respiratory irritants and CO is possible for initial attack crews on thefireline, primarily from peak smoke exposure situations.

Respiratory irritants—Exposure to the individual respiratory irritants (acrolein,formaldehyde, and PM3.5) was slightly greater than at project wildfires (compare figs. 15 and 25). At initial attack wildfires, the total irritant exposure (Em) on firelinesexceeded the recommended TLVs for about 10 percent of the firefighters and wasabout 90 percent of the PEL (fig. 24). These results were consistent with our observa-tion that the urgency of initial attack to fight wildfires entices fire crews to work throughintense smoke exposures.

In spite of the relatively high smoke exposures on the fireline, the long periods that initial attack crews spend on-call between fire dispatches lowered their shift-averageexposures. Compare the shift-average irritant exposure data at project wildfires and

56

ExposureAssessment for InitialAttack Wildfires

initial attack wildfires (figs. 15 and 25). The initial attack crews had consistently lowershift-average irritant exposures than the firefighters at project wildfires. During thedays when we tracked the initial attack crews, fires rarely occurred until the afternoon.As shown in figure 50, initial attack firefighters at wildfires had so much unexposedtime in their shift that no shift-length irritant exposures exceeded the PELs, even afterrecalculating the irritant exposures to compensate for potentially biased aldehyde measurements.

Peak exposures to respiratory irritants at initial attack wildfires (table 11) were similarto those at project wildfires. The large number of peak exposure samples at initialattack wildfires provided enough replication to conclude that peak exposure situationscause overexposure to respiratory irritants in about 50 percent of the firefighters,based on Em STELs calculated by using the TLVs, although none of the samplesexceeded the current PELs. If our aldehyde data were actually biased low, adjustingthem upward by a factor of 2 to 3 may change this conclusion. Similar results arepossible at project wildfires, although not enough peak exposure samples wereobtained at project wildfires to evaluate the possibility.

Benzene—There was little exposure to benzene at initial attack wildfires. All the benzene exposures on the fireline were less than 0.045 ppm (fig. 18). It was apparentthat smoke alone is not likely to cause an overexposure to benzene at initial attackwildfires. In the benzene data from initial attack wildfires, three of the four firefighterswith the highest fire-average exposures were working as sawyers or swampers,engine captains, or in mobile attack. The fourth was working directly adjacent to thesawyers on a tightly spaced hand crew. The highest peak benzene exposure samplesat initial attack wildfires also were among firefighters conducting direct attacks ormobile attacks or the engine captains (table 11). As for project wildfires, these datashowed that working near gasoline engines caused the largest benzene exposures.None of the peak benzene exposures approached ACGIH STEL recommendations.

57

Figure 50—Shift- and fireline-average respiratory irritant exposure at initial attack wildf-ires, assuming biased aldehyde data.

Carbon dioxide—Levels of CO2 in the breathing zone were all much less than the5000-ppm OSHA PEL (fig. 19). We had about as many TWA CO2 exposure estimatesat initial attack wildfires as at project wildfires. The distribution was skewed at initialattack wildfires by a few exposures that were substantially higher than the rest. Thispattern contrasted with results from project wildfires (fig. 9), where the exposures weremore evenly distributed across a narrower concentration range. Urban combustionsources possibly contributed to the higher CO2 levels at these initial attack wildfires.

Carbon monoxide—Only one firefighter had a fire-average CO exposure exceedingthe recommended TLV (fig. 20). As well, no firefighter had a shift-average CO expo-sure that exceeded the PEL when adjusted by using equation (2). Fewer shift-averageexposure measurements were obtained at initial attack wildfires. If the true distributionof exposures is such that only a small percentage of them exceeds the TLV, the appar-ent compliance with the TLV may be an artifact of a small sample size. Shift-averageoverexposures to CO probably occur on extended initial attacks, but clearly, the higherCO exposures that we found on the fireline often were diluted by the large proportionof time between fire dispatches.

Total suspended particulate—It is interesting that the total suspended particulateexposures at initial attack wildfires in southern California (fig. 23) were consistentlyhigher than at the project wildfires (fig. 13). The corresponding CO exposures explainthis result, because there is some correlation between the two pollutants in smoke(Sandberg and others 1989). The highest fireline-average CO exposure co-obtainedwith the TSP data was only 2.3 ppm at project wildfires compared to 11.8 ppm at theinitial attack wildfires. In fact, all the initial attack wildfire fireline CO exposures wereabove 4.8 ppm CO. The correspondingly higher TSP levels reinforced our conclusionsabout the potential TSP results for project wildfires and the need to further assess thishazard.

As is shown in the CO data, the initial attack TSP samples were all from firefighterswho had moderate smoke exposures. The highest TSP exposures on the fireline wereonly 60 percent of the OSHA PEL, and the TSP exposure levels we measured werenot significant when averaged over the work shift. Further sampling in a variety of fueltypes and lengths of initial-attack assignments might result in the occasional shift-average overexposure.

Our observations showed that overexposure to smoke is most likely to occur whenfirefighters are required to accomplish a task in spite of potential smoke levels andwhen weather or fire behavior causes smoke to be brought to the firefighter in highdoses. For example, a job task such as direct attack is inherently urgent, causingmany firefighters to ignore the irritation from smoke exposure as they focus onresponding successfully to the emergency. Some weather conditions such as strong,gusty, or erratic winds contribute to smoke exposure by causing unanticipated firebehavior or transporting smoke across firelines. A firefighter’s position (uphill or down-hill) and direction (upwind or downwind) relative to a fire are obvious factors contribut-ing to smoke exposure potential. Proving these concepts with statistical certainty isdifficult with the available data, but some patterns were apparent in the data. We wereable to evaluate two main factors thought to control smoke exposure potential: job taskand windspeed.

58

Factors in SmokeExposure

Job task, project wildfires—The job of putting out a large wildfire entails differenttasks. Our exposure measurements point to some differences in exposure amongthese work activities, but these differences were not proven to be statistically signifi-cant. There were not enough samples from many of the tasks to represent the associ-ated range of exposures (table 12), and further sampling will be required to evaluatedifferences. For many samples, other factors such as wind, fire behavior, or firefighterpositions relative to the fire seem to have defined the exposure during the samplemore than the work task itself. For example, a strong atmospheric inversion occurredone morning at the County Line fire (fire 20), during which all the sampled firefightersworked in very smoky conditions regardless of their task. In this obvious instance,these data were excluded from the work-type exposure comparison. Statistics aside,a few differences among work activities that were apparent to us are discussed below.

Digging fireline—Building firelines was a relatively low-exposure activity in mostinstances, as indirect firelines often were placed far from actively burning areas.When a direct fireline was constructed adjacent to rapidly burning fuels (for example,at fire 57) much higher exposures were sometimes observed. This scenario over-lapped what was categorized as “direct attack.” More often, the direct fireline con-struction that we observed was either around blackened areas where few hotspotswere actively burning or on the upwind or downhill edge of burning areas, where mostsmoke was carried away from the firefighters. In those cases, only intermittent periodsof smoke exposure occurred, with the overall average of moderate smoke exposureoccurring while firefighters were digging line (see for example figs. 26 and 29).

Holding fireline—Firefighters at project wildfires generally had the lowest exposureswhile “holding” firelines. At large project wildfires, “holding” personnel often wereassigned to a division or section of line with the task of being on guard for flare-ups orruns by the fire in that area. Because they were essentially waiting for something tohappen, their smoke exposures were usually low, as is apparent from figures 26 and27, where the asymmetrical shapes of the boxes and whiskers show that most of theholding personnel had low smoke exposures and only a few had higher exposures.These higher exposures (when something happened!) were due to wind-driven smokeas the fire tested the staffed section of fireline. Earlier results from prescribed burnsshow that holding personnel are among the more highly exposed groups (Reinhardtand others 1994), but the key difference was that the holding forces at prescribedburns were more often in a situation where the fireline was tested. At the project wildfires, holding personnel were seldom as near to actively burning fuels. However,as can be seen by the high exposures during the holding and mop-up job tasks (forexample, fig. 26), holding forces at project wildfires do endure the higher exposuresmeasured at prescribed fires. Most of the holding and mop-up job task data wereobtained from a single crew at the Libby complex fires in Libby, Montana, in 1994.This contract fire crew was assigned to hold a fireline during an afternoon burnoutoperation aimed at removing the hazard from a large area of unburned forest withinthe main firelines. Figure 51 shows a plot of the equivalent respiratory irritant levelscalculated by using equation (9) and the CO dosimeter data from one firefighter. Thelocal winds increased midway through the burnout, and smoke exposures increasedduring the afternoon as the winds transported smoke across the firelines.

59

The smoke exposure among holding forces can easily be three times the ACGIH recommended peak exposure limits for respiratory irritants and CO. At such levels,most workers will have adverse short-term health effects. Our limited sampling ofwildfires did not have a chance to further measure these conditions.

Lighting—Exposure to smoke during lighting was higher than we might have expect-ed, based on data for the task obtained during prescribed burns, where the exposureswere low for all pollutants except benzene (Reinhardt and others 1994). All the sam-ples of this job task were obtained during a burnout at fire 64 (Ann fire, Hamilton,Montana, 1994), where very erratic winds contributed to the smoke exposure. Suchconditions can occur at any time, but we believe that further sampling from a variety of wildfires would indicate a lower average exposure to CO and respiratory irritantsduring lighting. Benzene exposures may remain relatively high in any case becausethe drip torch is a constant exposure source during this task.

Mop-up—We found that particulate matter (total and respirable) were the only pollu-tants likely to be a significant health hazard during mop-up, with total suspended particulate comprising the greatest concern with an average of nearly half the PEL(fig. 30). These results were intuitive, because the task involves digging and stirring ofashes and dirt, which cause particulates to become airborne. Smoke exposure duringmop-up is the best characterized activity at project wildfires. With the exception ofTSP, we are confident that the distribution of smoke exposure data for mop-up at proj -ect wildfires accurately represented what most firefighters experience. From a healthconsequence standpoint, the respirable particulate exposure is the main concern,averaging over half the ACGIH recommended TLV of 3.0 mg/m3 (fig. 29). One goal for future monitoring efforts could be to augment the total suspended particulate databecause our nine samples are inadequate to define the exposure potential.

60

Figure 51—Respiratory irritant exposure from one firefighter as that person was holding the fireline during a burnout operation.

Sawyer—Smoke exposure among sawyers at project wildfires was not especially highfor any pollutant except benzene. One 93-minute benzene sample result from fire 64was 0.82 ppm, which is above ACGIH recommended TLV but below the PEL. Becausethe other pollutants were at low concentrations during the sample period, yet they nor-mally are well correlated to benzene in smoke, this high benzene sample was mostlikely due to gasoline exposure. Sawyers, pump operators, and fueling personnelshare this potential for greater benzene exposures.

The 13 exposure samples obtained from sawyers at project wildfires did not documentelevated CO exposures. We expected to see greater CO levels among those workingso near to the exhaust of a chainsaw, especially if ambient winds were low or firelineswere being constructed in dense fuels, such as chaparral, so that the engine exhaustwas not well dispersed. The sawyers that we monitored were not in such conditions atthe project wildfires. The same comments apply to swampers, and our three exposuresamples from swampers at project wildfires did not show significantly elevated COexposures.

Direct attack—The formaldehyde data provided a good indication of smokeexposures during direct attack relative to the other work activities (fig. 27). The 11exposure samples from this task indicated slightly higher formaldehyde exposure thanthe other tasks. The smoke samples from direct attack at project wildfires were notobtained in especially smoky conditions, but for a given firefighter they showed highersmoke concentrations during the direct attack of spot fires and while digging line in adirect attack action than during bracketing periods of mop-up or line construction.

Engine—At project wildfires, all smoke exposure samples for the “engine” job taskwere obtained from a firefighter assigned to tend a portable gas-powered pump serv-ing a fireline. This firefighter had minor exposure to drifting smoke from the fire, butremained at the road away from the fire for the entire work shift. We surmise that hisrelatively high CO and benzene exposures were mostly from the pump exhaust andexhaust from passing vehicles. The relatively high benzene levels were consistentwith gasoline fumes. One 155-minute CO sample averaged 27 ppm, which under-scored the point that workers may not always be aware of the hazards of working near operating engines.

Job task, initial attack wildfires—As at project wildfires, smoke exposure seemed todiffer among firefighters depending on their job task. The differences were not statisti -cally significant. More samples would be needed to evaluate statistical significance.

Attack—Smoke exposure among firefighters at initial attack wildfires reached thehighest levels during direct attack and mobile attack and while digging line in a directattack action (table 13). The average exposure to CO during these activities wasabout half the ACGIH recommended TLV, and respiratory irritant exposures averagedabout 85 percent of the ACGIH recommended TLV standard. The highest exposuresto both were above the ACGIH recommended TLVs, but few firefighters spent morethan 8 hours on the fireline. Considering the variability of the exposures during thesethree tasks, adverse short-term health effects were expected, especially among work-ers sensitive to smoke.

61

The average smoke exposures during “attack” activities were somewhat higher at ini-tial attack wildfires than at project wildfires. Three explanations may apply: (1) at smallfires, firefighters often work in closer proximity to actively burning areas with less riskof entrapment than is possible at project wildfires and thus may receive higher smokeexposures; (2) fire crews during initial attack may endure brief, intense smoke expo-sures because they are in an emergency-response situation with the opportunity tocontrol an incipient fire before more valuable resources are destroyed; and (3) relative-ly fewer direct attack samples were available from project wildfires and we may nothave captured the range of exposures during this activity. A well-planned, routinesmoke exposure monitoring program could show whether smoke exposures duringdirect attack differ between large and small fires.

Mop-up—Smoke exposure among firefighters during mop-up was low relative to theother work activities at initial attack wildfires, averaging about 45 percent of AGCIHrecommended TLV for respiratory irritants (table 13). Respiratory irritant exposuresamong firefighters was similar at both initial attack wildfires and project wildfires, butthe CO exposures were somewhat higher at the initial attack events (compare figs. 26and 31).

Engine—The engine job category at initial attack wildfires included wildland fireengine operators and captains, rather than the pump operators represented at projectwildfires. The initial attack “engine” individuals had low to moderate smoke exposuresrelative to the other work activities. They were either driving the fire engines duringmobile attack or operating their engine’s pumps and directing their crew’s efforts.

Sawyer and swamper—The sawyers and swampers at the initial attack wildfires wereworking at close quarters within their hotshot crews to construct fireline adjacent toburned areas. Their CO exposures were greater than those for individuals performingthe same tasks at project wildfires. As discussed above, the ability to work closelywith small fires may cause higher smoke exposures at initial attack wildfires. Anotherfactor in our data may be that the sawyer-swamper teams at initial attack wildfireswere working either in dense brush (fires 66 and 70) or a narrow draw (fire 66); bothsituations could increase worker exposure by limiting the dispersion of exhaust fromchainsaws.

Wind speed—Smoke exposure among firefighters engaged in mop-up showed notrend relative to ambient wind speed, but a correlation with wind was apparent amongthose working in “attack” activities. Figure 37 shows the relation. Smoke exposurerose by about 1 ppm CO with each additional mile per hour of windspeed. The r2 forthe regression was only 0.23, which limited its value as a predictive tool, but the corre-lation supports the observation that firefighters encounter smokier conditions near thehead of the fire when the winds are strong. This result might be expected because fireintensity and rate of spread increases with ambient windspeed, as does the difficultyof suppressing the fire. Most of the samples represented in this graph were from initialattack wildfires, because wind data were not available for the project wildfires. A simi-lar correlation may exist at project wildfires, however, because the highest sustainedsmoke exposures we measured at those occurred when an increasing afternoon windtransported smoke from a burnout operation into a crew of firefighters (fire 61, dis-cussed above under holding and mop-up exposures at project wildfires). In future

62

studies, data about whether a firefighter was upwind or downwind of a fire would helpto clarify the role of windspeed: a high wind does not increase smoke exposure whenfirefighters are upwind of the fire.

Atmospheric inversions that limit dispersion of smoke provide another set of conditionswhere widespread and consistently high smoke levels are possible (as evidenced byseveral exposures at 35 to 55 ppm CO measured over 3 hours in such conditions atfire 20). Similar conditions are reported to have occurred at the Happy Camp wildfirecomplex in the Klamath National Forest in 1987 (Sutton and others 1988).

Other site factors—We believe that site factors are likely to exert influence on smokeexposure, but our data were inadequate to test this. Combustion research has shownthat fires in heavier or wetter fuels produce more pollutants per ton of fuel consumedthan those in dry or light fuels, which burn more completely (Ward and Hardy 1991).Because fuel moisture is a key determinant of fire behavior and intensity, a fire in fuelswith high moisture content will burn more slowly and less completely, thereby increas-ing pollutant production. If such a slow-moving fire must be approached closely orfrom the downwind side, then firefighters would seem likely to have significant smokeexposures.

Differences in smoke exposure may exist among firefighters working in different fueltypes due to differences in emission factors for the pollutants and because of differenttactics employed for the specific fuel. For example, firefighters working with tractorand plow equipment in grass and brush in the Southeastern United States may havevery different smoke exposure patterns than hand crew personnel working in steeptimber in the Western United States. We do not have enough data from other regionsand diverse vegetation types to evaluate these factors.

The electronic data-logging dosimeter is the best way to collect widespread CO exposure data. Our decision to rely exclusively on electronic data loggers in the 1995season was based on our comparisons of the CO results from our reference method(ICM 128) with those obtained by the dosimeters during 1993-94 (fig. 38). Althoughthe dosimeter data were slightly lower than those from the reference method, we wereconfident that rigorous quality assurance could overcome this problem. The dosime-ters tended to become less sensitive to CO over a work shift (fig. 39), but the occur-rence of this bias could be tested by presampling and postsampling QC checks of thesensor accuracy. Adhering to strict QC limits for the amount of acceptable bias yield-ed CO exposure data that are comparable among groups and methods. Figures 38and 39 suggest that dosimeter data obtained without frequent objective checks ofinstrument performance are considered estimates. Such data would be of no value to a long-term, routine smoke exposure monitoring program.

Our QA program included evaluation of matrix effects on the data-logger sensors arising from interferences from the other components of smoke. By spiking knownamounts of CO into whole-air samples of smoke, we showed that the accuracy of theDraeger dosimeter was not unduly affected by the smoke matrix at the levels we rou-tinely monitored (fig. 40). We suggest that such an evaluation occur before a dosime-ter model is selected for routine use at fires.

63

DosimeterPerformance

After our experience with the Draeger 190 data loggers, we concluded that the manu-facturer’s recommendations for calibration frequency were insufficient to obtain ade-quate accuracy in wildland fire conditions. We suggest twice-daily QC checks and acalibration frequency based on the QC check results. We also observed temperaturesensitivity in the Draeger 190 data logger. Placing a dosimeter in direct sunlight on a hot day caused the data logger to falsely indicate up to 5 ppm CO; moving thedosimeter to the shade corrected the problem. The Draeger data loggers we evaluat-ed were highly susceptible to water contacting the inlet filter and disabling the instru-ment, a consideration when evaluating product suitability to the rugged field conditionsof firefighting. Radio-frequency shielding is another beneficial feature of many models.There are now many brands of electronic dosimeters available, with and without data-logging capability. With appropriate QC programs in place, most brands are capableof providing reliable information to the user.

A beneficial aspect of data-logging dosimeters is the ease of data collection and therich information content of the exposure profiles, which provide continuous traces ofCO levels throughout the monitoring period (refer to fig. 51). Detailed observationsand timekeeping allow for postexposure debriefings on the smoke exposures that afire crew encounters. Printing the graphs and data summaries provides permanentrecords of CO exposure, and the transfer of data to personal computers is simple. Wewere even able to accomplish this in the field at remote spike camps by using a note-book computer and a small 9-volt battery to power the Draeger data transfer adapter.With two small gas cylinders (a calibration gas and a QC check gas) and a portablecomputer, monitoring equipment adequate for weeks of data collection among severalcrews can weigh less than 3.6 kilograms and be as mobile as any fireline situationrequires.

Most importantly, dosimeters provide the alarm capability to warn users of unhealthfulCO levels and provide crew foremen or safety officers with an objective indicator ofsmoke intensity. Even though the combination of respiratory irritants reaches criticallevels for crew health before CO does, CO can be monitored and related to the sum of respiratory irritants by a simple formula. By relying on dosimeters for a monitoringprogram, fire managers can decide objectively when action should occur, such as donning respiratory protection against irritants or evacuating when CO becomes hazardous.

Assume, for example, that the TWA CO exposure of a firefighter at a project wildfire is23 ppm. This is within the AGCIH recommended exposure limit of 25 ppm. From theinterpollutant regression equations in table 7, the corresponding irritant exposures areas follows:

TWA exposure regression Recommended exposurePollutant result limit––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––Formaldehyde (HCHO) 0.087 ppm 0.3 ppmRespirable particulate

(PM3.5) 1.94 mg/m3 3 mg/m3

Acrolein (ACRO) 0.010 ppm 0.1 ppm

64

The combined irritant exposure is derived as follows from equation (1):

Em =[HCHO]

+[PM3.5]

+[ACRO]

–––––– –––––– ––––––0.3 3 0.1

Em =[0.087]

+[1.94]

+[0.01]

––––– –––– ––––0.3 3 0.1

Because the irritant exposure exceeds 1.0, it does not meet AGCIH irritant exposurelimits, and mitigating measures may need to be implemented to fall within compliance.At 23 ppm, the firefighter was not overexposed to CO. By wearing lightweight respira-tory protection, the firefighter could continue to work without incurring harm from respi-ratory irritation. With proper advance training in respirator use, a lightweight respiratorfor fine particles could achieve this irritant control, providing CO is monitored closely.Carbon monoxide monitoring is important when using respirators because they do not protect against CO exposure even though they afford relief from respiratory tractirritation.

Observers’ estimates of smoke exposure were variable but still seemed to be a practi-cal tool for exposure management in the absence of real data. Based on our limiteddata, we believe that with further data from a broader range of exposures, observers’estimates may be widely useful for smoke exposure management. In situations suchas those depicted in figures 41 through 45, our observers classified the actual expo-sure fairly well (see figs. 46 through 48). Although there is variation in the results, wefound that a smoke class of 3 (medium) related to a CO exposure of about 30 ppm.Based on this, it is suggested that firefighters limit their work in “medium” smoke tokeep smoke exposures within recommended limits. The regression for PM3.5 was the best among those examined, with an r2 of 0.6 for 46 pairs of observer estimatesand PM3.5 samples. The regressions for CO and formaldehyde were not as precise,probably because only particulate matter reflects, scatters, or absorbs visible light—alleffects detectable by eye. Not enough data were obtained for total suspended particu-late or acrolein to provide useful regressions, and none of the regressions containsenough data for high concentrations to determine the utility of this approach in verysmoky conditions. For fire managers without access to monitoring equipment, thesmoke classifications provide an interim approach to assess and control smokeexposure.

If the goal of managers is to minimize the exposure of firefighters to unhealthful levelsof smoke that could exceed legal and recommended limits during wildfires, then man-agers could implement a program to manage smoke, composed of the following ele-ments: (1) training in hazard awareness, (2) monitoring of routine smoke exposure,(3) evaluating exposure limits, (4) improving recordkeeping, (5) assessing health risks,(6) deploying respirators, and (7) involving workers and regulators.

The data we have indicate that firefighters mostly work in smoke levels that are notexpected to cause health problems or exceed legal and recommended limits. Existingwork practices are adequate to protect firefighters’ health in those situations. Our datashow, though, that firefighters occasionally work in situations where they enduresmoke levels that exceed guidelines recommended by occupational health experts,

65

Observer SmokeEstimates

ManagementImplications

levels that can even be higher than U.S. occupational safety regulations allow.Reducing such overexposures will reduce organizational liability and is expected toincrease work force productivity and health. Wildland fire agencies could considercontrolling these high-exposure situations through a focused program to reduce overexposure to smoke. Wildland fire agencies such as the Forest Service, BLM, andCDF already have in place many of the elements of a successful program. With minorrevisions, development of pieces missing from the strategy, and appropriate technicaland administrative oversight, overexposures to smoke can be reduced. The basic outline of a smoke exposure management program follows.

For change to occur, there must be consensus that change is needed. The work forcemust understand the hazards of smoke exposure, the short- and long-term healtheffects, the situations that cause overexposure to smoke, and methods to assess andavoid overexposure. The education must begin in the classroom, before firefightersare assigned to the field, but should be routinely refreshed to maintain a positivesafety attitude. Field education about the hazards of firefighting is also important, andsmoke hazards should be reviewed when they occur. The visual smoke exposureclassifications we developed are a simple start to heighten worker and managerawareness of smoke exposure on the fireline, but it is important to include objectiveCO dosimetry and routinely inform firefighters of monitoring results.

Data gathered widely in a manner well integrated with the incident command systemwill allow adequate tracking of existing conditions and future trends in smoke expo-sure. Personal smoke exposure assessment could be added to the objectives of wild-fire and prescribed fire safety management planning. A cache of equipment could beobtained, and safety officers or other personnel trained and held accountable forimplementation of a well-designed monitoring plan at some representative percentageof fires. The sampling plan could achieve regional balance and capture smoke expo-sure data at initial attacks, during extended attack at project wildfires, and at pre-scribed burns. The program could be based on CO data loggers, after evaluation ofthe available products to select the best suited dosimeters. The monitoring could bedone with strict adherence to protocol so that all data are comparable and of knownquality. A small percentage of the monitored crews could be randomly selected for further evaluation of the correlations between CO and the respiratory irritants, andTSP could be monitored widely until that hazard is adequately assessed. The datacould be evaluated at least annually to detect trends in exposure and refine exposuremanagement strategies. The program could include periodic third-party data evalua-tion to validate data quality. A well-designed program can be done with little adverseimpact on fire operations, provide many benefits, and cost surprisingly little.

An independent panel of expert toxicologists could be convened to evaluate exposuredata and the potential health effects among firefighters. This evaluation process couldbe used to determine whether the existing OSHA PELs are adequate or whether alter-native standards should be derived. Consensus standards can then be set. The wild-land fire workplace is very different from an industrial or office environment, and limitsappropriate for firefighting may differ from those taken “off-the-shelf.” Depending onthe pollutant, current exposure limits may be inappropriate.

66

Training in HazardAwareness

Monitoring of RoutineSmoke Exposure

Evaluating ExposureLimits

The existing accident reporting system at wildland fires does not enable efficient andconfidential tracking of smoke-related illnesses among fireline personnel. Computerscould enable better tracking of all work-related injury data to ultimately provide epi-demiological data for those involved with the long-term health of the work force. Goodrecordkeeping is needed to measure progress in meeting injury and illness reductiongoals.

Risk assessment is a valuable tool to evaluate the long-term health risks to firefight-ers from smoke exposure. A preliminary assessment has been done (Booze andReinhardt 1996), but a refined risk assessment would be helpful if it included exposuredata for all pollutants that firefighters face, including entrained dust exposures. Thiswould improve upon existing toxicological dose-response data for fine particles anduse realistic exposure assumptions.

Our exposure data for respiratory irritants indicated that exposure to these may needto be reduced in peak exposure situations. Respirators are an effective way to do this.Most firefighters will seldom need to wear a respirator and may need them only forbrief periods to mitigate the hazards. By donning respiratory protection on thoseoccasions when the acute hazards would be reduced (except CO, with present tech-nology), the firefighters still could be effective at their jobs—perhaps more so if the irritants were becoming incapacitating. Even though current respirator technology isnot ideal, it is light and comfortable enough to enable firefighters to protect their lungsand maintain productivity in irritating and possibly damaging conditions.

Because respirator use is regulated by OSHA and equivalent state agencies, respira-tor deployment should occur only in the context of a respiratory protection program.Key elements in these programs include assessing the ability of individuals to workwith respirators, training employees on the limits of respirator protection, and describ-ing the implementation of medical surveillance procedures. Our data also suggestthat current respirators for organic vapors and particulates should be worn in conjunc-tion with CO alarm dosimeters to warn of concomitant CO overexposure. This is onlyan interim solution. New respirator technologies are needed to protect against COexposure as well. Such a respirator design once was used by urban firefighters butwas banned after being implicated in structural firefighting deaths. Because wildlandfire atmospheres have not been shown to be immediately dangerous to life and health(without risking severe burns), it makes sense to develop a new lightweight respiratorthat protects against CO as well as the irritants. The main barrier is the lack of warning properties for CO overexposure. Reliable service-life indicators could bedeveloped, however, for respirator cartridges to afford adequate protection for wildlandfirefighters.

Regular dialogue with OSHA or the responsible state agency, as well as employeerepresentatives, is a key to ensuring that the smoke exposure management plan isworkable and meets consensus goals. If the plan is not accepted by workers andmanagers, then it will not succeed. Finally, the regulatory agencies could be consult-ed for interpretations and clarifications on legal aspects and early review of the smokeexposure management plans.

67

ImprovingRecordkeeping

Assessing HealthRisks

DeployingRespirators

Involving Workers andRegulators

Our study found that smoke exposure among firefighters is a problem that is not widespread but does warrant management action to improve current conditions. Webelieve we have identified the conditions associated with overexposure to smoke andhave found that they are amenable to control through carefully designed managementstrategies. By implementing a smoke exposure management plan, fire managementagencies will enhance their wor kers’ health and productivity.

ACGIH American Conference of Governmental Industrial Hygienists

BLM Bureau of Land Management

C Ceiling limit

C3H4O Acrolein

C6H6 Benzene

CDF California Department of Forestry and Fire Protection

CFK Coburn-Forster-Kane

CO Carbon monoxide

CO2 Carbon dioxide

COHb Carboxyhemoglobin

CS2 Carbon disulfide

DNPH 2,4-dinitrophenylhydrazine

Em Equivalent exposure limit for a mixture

EPA U.S. Environmental Protection Agency

HCHO Formaldehyde

HPLC High-performance liquid chromatography

ICM Intersociety Committee Method

MDL Method detection limit

mg/m3 Milligrams per cubic meter

µm Micrometer

MTDC Missoula Technology Development Center

NFDRS National Fire Danger Rating System

NIOSH National Institute for Occupational Safety and Health

NIST National Institute of Standards and Technology

NPS National Park Service

68

Abbreviations

NWCG National Wildfire Coordinating Group

OSHA Occupational Safety and Health Administration

PE Performance evaluation

PEL Permissible exposure limit

PM3.5 Respirable particulate matter

ppm Parts per million

QA Quality assurance

QAPP Quality assurance project plan

QC Quality control

r2 Coefficient of determination

SOP Standard operating procedure

STEL Short-term exposure limit

TLV Threshold limit value

TSP Total suspended particulate

TWA Time-weighted average

USDA U.S. Department of Agriculture

We wish to acknowledge the support and cooperation of the Pacific Northwest, PacificSouthwest, and Northern Regions, U.S. Department of Agriculture, Forest Service;the U.S. Department of the Interior, Bureau of Land Management and National ParkService; the National Wildfire Coordinating Group; the State of California, Departmentof Forestry and Fire Protection; and Dr. Bjorn Hrutfiord and the University ofWashington Department of Forestry.

Outstanding personal contributions to this research were made by Aimee Backlund,Jenelle Black, Ray Borgen, Todd Burke, Tanya Copeland, Stacey Drury, DavidFrewing, Andrew Hanneman, Michael Hallett, Ken Harris, Brian High, Cheryl Luschei,Alexis Merydith, Doug Milligan, Phil Monsanto, Charles Sauvageau, KathySiebenmann, and Bob Vihnanek.

American Conference of Governmental Industrial Hygienists. 1996. 1996 TLVs®

and BEIs®: Threshold Limit Values for chemical substances and physical agents and biological exposure indices. Cincinnati, OH. 138 p.

Booze, T.F.; Reinhardt, T.E. 1996. Assessment of the health risks of chronic smokeexposure for wildland firefighters. [Place of publication unknown]: [Publisherunknown]. Unpublished report. Available from: Radian Corporation, 8501 N. MopacBlvd., Austin, TX 78759.

69

Acknowledgments

References

Deeming, J.E.; Burgan, R.E.; Cohen, J.D. 1977. The national fire-danger rating sys-tem—1978. Gen. Tech. Rep. INT-39. Ogden, UT: U.S. Department of Agriculture,Forest Service, Intermountain Forest and Range Experiment Station. 63 p.

Dost, F.N. 1991. Acute toxicology of components of vegetation smoke. In: Reviews of environmental contamination and toxicology. New York: Springer-Verlag: 2-46.Vol. 119.

Harrison, R.; Materna, B.; Liu, D. [and others]. 1992. Respiratory effects of smokeexposure in wildland firefighters: methacholine challenge testing and exposuremonitoring. [Place of publication unknown]: [Publisher unknown]. 30 p. Availablefrom: California Occupational Health Program, California Department of HealthServices, 2151 Berkeley Way, Rm. 504, Berkeley, CA 94704.

Jackson, G.;Tietz, J.G. 1979. Preliminary analysis: firefighters’ exposure to carbonmonoxide on wildfires and prescribed burns. Project Record. Missoula, MT: U.S.Department of Agriculture, Forest Service, Equipment Development Center.

Letts, D.; Fidler, A.T.; Deitchman, S. [and others]. 1991. Health hazard evaluationreport, U.S. Department of the Interior, National Park Service, southern California.HETA 91-152-2140. Cincinnati, OH: U.S. Department of Health and HumanServices, Public Health Service, Centers for Disease Control, National Institute for Occupational Safety and Health. 16 p.

Liedel, N.A.; Busch, K.A; Lynch, J.R. 1977. Occupational exposure sampling strategy manual. Cincinnati, OH: U.S. Department of Health, Education andWelfare, National Institute for Occupational Safety and Health: 125.

Lodge, J.P., ed. 1989. Method 128. In: Methods of air sampling and analysis. 3d ed.Chelsea, MI: Lewis Publishers, Inc.: 296-302.

Materna, B.L.; Jones, J.R.; Sutton, P.M. [and others]. 1992. Occupational exposuresin California wildland fire fighting. American Industrial Hygiene AssociationJournal. 53 (1): 69-76.

National Institute for Occupational Safety and Health. 1989a. Manual of analyticalmethods. 3d ed. Cincinnati, OH: U.S. Department of Health and Human Services,Public Health Service, Centers for Disease Control: 0500-1 to 0500-3.

National Institute for Occupational Safety and Health. 1989b. Manual of analyticalmethods. 3d ed. Cincinnati, OH: U.S. Department of Health and Human Services,Public Health Service, Centers for Disease Control: 0600-1 to 0600-6.

National Institute for Occupational Safety and Health. 1989c. Manual of analyticalmethods. 3d ed. Cincinnati, OH: U.S. Department of Health and Human Services,Public Health Service, Centers for Disease Control: 1501-1 to 1501-7.

Neter, J.; Wasserman, W; Kutner, M.H. 1983. Applied linear regression models.Homewood, IL: Irwin Publishers: 111-122.

70

Radian Corporation. 1993. Smoke exposure assessment quality assurance projectplan. Austin, TX.

Radian Corporation. 1996. Smoke exposure assessment at wildfires and prescribedburns quality assurance project plan. Austin, TX.

Reh, C.M.; Deitchman, S.D. 1992. Health hazard evaluation report, U.S. Departmentof the Interior, National Park Service, Yellowstone National Park, Wyoming. HETA88-320-2176. Cincinnati, OH: U.S. Department of Health and Human Services,Public Health Service, Centers for Disease Control, National Institute forOccupational Safety and Health.

Reh, C.M.; Letts, D.; Deitchman, S. 1994. Health hazard evaluation report, U.S.Department of the Interior, National Park Service, Yosemite National Park,California. HETA 90-0365-2314 (2415).Cincinnati, OH: U.S. Department of Healthand Human Services, Public Health Service, Centers for Disease Control,National Institute for Occupational Safety and Health. 33 p.

Reinhardt, T.E.; Black, J.; Ottmar, R.D. 1995a. Smoke exposure at northernCalifornia vegetation fires: [Place of publication unknown]: [Publisher unknown];final report. Available from: U.S. Department of Agriculture, Forest Service, PacificNorthwest Research Station, 4043 Roosevelt Way NE, Seattle, WA 98105.

Reinhardt, T.E.; Black, J.; Ottmar, R.D. 1995b. Smoke exposure at Pacific North-west wildfires. [Place of publication unknown]: [Publisher unknown]; final report.Available from: U.S. Department of Agriculture, Forest Service, Pacific NorthwestResearch Station, 4043 Roosevelt Way NE, Seattle, WA 98105.

Reinhardt, T.E.; Hanneman, A.H.; Ottmar, R.D. 1994. Smoke exposure at prescribedburns: [Place of publication unknown]: [Publisher unknown]; final report. 123 p.Available from: U.S. Department of Agriculture, Forest Service, Pacific NorthwestResearch Station, 4043 Roosevelt Way NE, Seattle, WA 98105.

Reinhardt,Tim; Ottmar, Roger D. 1997a. Smoke exposure at western wildfires.[Place of publication unknown]: [Publisher unknown]; final report. 149 p. Availablefrom: U.S. Department of Agriculture, Forest Service, Pacific Northwest ResearchStation, 4043 Roosevelt Way NE, Seattle, WA 98105.

Reinhardt,Timothy E.; Ottmar, Roger D. 1997b. Smoke exposure among wildlandfirefighters: a review and discussion of current literature. Gen. Tech. Rep. PNW-GTR- 373. Portland, OR: U.S. Department of Agriculture, Forest Service, PacificNorthwest Research Station. 61 p.

Rothman, N.; Ford, D.P.; Baser, M.E. [and others]. 1991. Pulmonary function andrespiratory symptoms in wildland firefighters. Journal of Occupational Medicine.33(11): 1163-1169.

71

Sandberg, D.V.; Ward, D.E.; Ottmar, R.D. [and others]. 1989. Mitigation of prescribedfire atmospheric pollution through increased utilization of hardwoods, piledresidues, and long-needled conifers. [Place of publication unknown]: [Publisherunknown]; final report. 160 p. Available from: U.S. Department of Agriculture,Forest Service, Pacific Northwest Research Station, 4043 Roosevelt Way NE,Seattle, WA 98105.

SAS Institute, Inc. 1989. SAS/STAT® user’s guide, version 6, 4th edition. Cary, NC.846 p. Vol. 2.

Shapiro, S.S.;Wilk, M.B. 1965. An analysis of variance test for normality (completesamples). Biometrika. 52: 591-611.

Smith, S.R.; Steinberg, S; Gaydos, J.C. 1996. Errors in derivations of the Coburn-Forster-Kane equation for predicting carboxyhemoglobin. American IndustrialHygiene Association Journal. 57: 621-625.

Steel, R.G.D.;Torrie, J.H. 1980. Principles and procedures of statistics—a biometricalapproach. 2d ed. New York: McGraw Hill. 633 p.

Sutton, P.; Castorina, J.; Harrison, R.; Ford, D.P. 1988. Carbon monoxide exposurein wildland fire fighters. Field Investigation FI-87-008. [Place of publicationunknown]: [Publisher unknown]. Available from: California Department of HealthServices, Epidemiologic Studies and Surveillance Section and OccupationalHealth Surveillance and Evaluation Program, 2151 Berkeley Way, Rm. 504,Berkeley, CA 94704.

Tejada, S.B. 1986. Evaluation of silica gel cartridges coated in-situ with acidified 2,4-dinitrophenylhydrazine for sampling aldehydes and ketones in air. InternationalJournal of Analytical Chemistry. 26: 167-185.

U.S. Department of Labor. 1979. Compliance officer’s field manual. Washington, DC:Occupational Health and Safety Administration.

U.S. Department of Labor. 1993. Carbon monoxide in workplace atmospheres(direct-reading monitor). Method No. ID-209. Salt Lake City, UT: OccupationalSafety and Health Administration, Technical Center. 49 p.

U.S. Environmental Protection Agency 1984. Code of federal regulations, title 40,chapter I, part 136, appendix B: definition and procedure for the determination ofthe method detection limit–revision 1.11. Washington, DC: 198.

U.S. Environmental Protection Agency, Research and Development. 1986.Supplement to EPA/600/4-84/041: compendium of methods for the determinationof toxic organic compounds in ambient air. Research Triangle Park, NC:Environmental Monitoring Systems Laboratory. 38 p.

Ward, Darold E.; Hardy, Colin C. 1991. Smoke emissions from wildland fires.Environmental International. 17: 117-134.

72

This page has been left blank intentionally.Document continues on next page.

This page has been left blank intentionally.Document continues on next page.

The Forest Service of the U.S. Department of Agriculture is dedicat-ed to the principle of multiple use management of the Nation’s forestresources for sustained yields of wood, water, forage, wildlife, andrecreation.Through forestry research, cooperation with the Statesand private forest owners, and management of the National Forestsand National Grasslands, it strives—as directed by Congress—toprovide increasingly greater service to a growing Nation.

The U.S. Department of Agriculture (USDA) prohibits discriminationin all its programs and activities on the basis of race, color, nationalorigin, gender, religion, age, disability, political beliefs, sexual orien-tation, or marital or family status. (Not all prohibited bases apply toall programs.) Persons with disabilities who require alternativemeans for communication of program information (Braille, large print, audiotape, etc.) should contact USDA’s TARGET Center at (202) 720-2600 (voice and TDD).

To file a complaint of discrimination, write USDA, Director, Office ofCivil Rights, Room 326-W, Whitten Building, 14th and IndependenceAvenue, SW, Washington, DC 20250-9410 or call (202) 720-5964(voice and TDD). USDA is an equal opportunity provider andemployer.

Pacific Northwest Research Station333 S.W. First Avenue P.O. Box 3890Portland, Oregon 97208-3890

U.S. Department of AgriculturePacific Northwest Research Station333 S.W. First AvenueP.O. Box 3890Portland, OR 97208

Official BusinessPenalty for Private Use, $300